# Quarrion — full content > Quarrion is an AI job-search agent: it scans 100+ job boards, scores every role against your CV, tailors your application, and drafts your recruiter replies — so you wake up to interviews. Every public page of this site, concatenated as markdown. The link index is at https://quarrion.ai/llms.txt. Each page is also available individually as markdown, either by requesting the canonical URL with `Accept: text/markdown` or by appending `.md` to it. The salary section states every published cell inline. Full per-cell methodology lives at that cell page with `.md` appended — for example https://quarrion.ai/salary/london/chef-de-partie.md. Every salary surface prints the same `Data generation:` token; if two of them differ, you are holding two different snapshots. --- # Quarrion — AI Job Search Agent That Applies For You Canonical URL: https://quarrion.ai > Quarrion is an AI job-search agent: it scans 100+ job boards, scores every role against your CV, tailors your application, and drafts your recruiter replies — so you wake up to interviews. ## Your job search, run by an agent. Quarrion scans, scores, tailors and applies — you interview. - Start free: https://quarrion.ai/signup - Sign in: https://quarrion.ai/login ## What the agent does - **DISCOVERS.** 100+ boards - **SCORES.** against your CV - **TAILORS.** CV + cover letter per role - **APPLIES.** forms filled for you - **REPLIES.** recruiter drafts ready ## Why it is different ### 1. Deep research, not keyword overlap A three-tier pipeline scores every role against your CV, then researches the ones worth your time and tags what it found. Category gap: Most tools compare keywords and call it a match score. ### 2. Dead listings filtered out Every listing is probed for liveness before it reaches you, so you are not applying into a closed req. Category gap: Ghost listings are the single most common complaint in this category. ### 3. Auto-apply with an approval gate The agent fills and submits real applications, but you approve what goes out. Quality without spray-and-pray. Category gap: Blasters get accounts bot-flagged; autofill tools still leave the work to you. ### 4. Salary intelligence that is real Published medians of what employers actually advertised — every sector, stated pay only, with the posting count on every figure. Plus city-level benchmarks adjusted for cost of living and inflation, with live FX, on an explorable globe. Category gap: Salary figures are often quoted with no sample behind them, or with missing pay filled in by a model. ### 5. Outreach and follow-ups built in A unified recruiter inbox with AI-drafted replies and follow-ups tied to the application record. Category gap: Trackers store contacts. They do not close the loop. ### 6. CV tailoring that cannot invent skills Tailoring is grounded in your own profile and knowledge graph, so it rewrites what you have rather than inventing what you do not. Category gap: Hallucinated skills are a documented problem with AI CV tools. ### 7. The whole pipeline in one place Discovery, scoring, research, tailoring, applying, messaging, interview prep, and offer negotiation. Category gap: Every alternative covers two or three of these stages. None covers the loop. ## What it looks like - Every role scored against your CV — ranked, not listed. (Top Picks board showing roles ranked by match score with research tags) - And it shows its working, so you can disagree with it. (An expanded role card showing the score breakdown and research findings) - Every application tracked from applied to offer. (Application pipeline board with roles arranged across stage columns) - Recruiter replies drafted and waiting for your approval. (Recruiter message thread with an AI-drafted reply ready to send) - What the market keeps asking for that you do not have yet. (Insights view showing skill coverage across matched roles) - And what the job is worth, wherever it is. (Interactive 3D globe showing salary benchmarks by city) Product tour: The Top Picks board filling with scored roles, each tagged with what the agent researched. ## Common questions ### Does it apply to jobs without asking me? No. The agent prepares each application and queues it for your approval — you decide what goes out. You can review every application before it is submitted. ### Do I need to upload a CV to start? Yes, and it is the only setup step that matters. Everything the agent does — scoring, tailoring, drafting — is grounded in your CV, so it will not invent experience you do not have. ### Is it actually free to start? Yes. The free tier needs no card and gives you scored jobs on your first session. Paid plans add volume, auto-apply and tailoring. ### What if I get hired? Then it worked, and you should cancel. You can cancel from your settings at any time and keep access until the end of the period you have paid for. ### Which job boards does it cover? Over 100 boards and company career pages, including Ashby, Workable, Greenhouse and Lever, plus aggregator feeds — with each listing checked to still be open before it reaches you. --- # Simple, transparent pricing Canonical URL: https://quarrion.ai/pricing > Start free. Upgrade when your search gets serious. Cancel anytime. 50% off for the first 100 users — limited time ## Plans ### Free Discover and AI-score jobs against your CV. No card required. Free — no card required. - 50 job discoveries/month - 25 AI scores/month - 5 follow-ups/month - 3 salary explorer lookups/month - 3 job boards - Basic analytics ### Starter For an active search: CV tailoring, cover letters, and auto-apply. £6.99/week · £24.99/month - 500 job discoveries/month - 200 AI scores - 20 deep research - 5 CV tailoring - 3 cover letters - 10 auto-apply - 3 interview prep - All job boards - Email outreach - Full analytics ### Pro The full pipeline — unlimited discovery, deep research, and salary negotiation. £14.99/week · £49.99/month - Unlimited job discoveries - 1000 AI scores - 100 deep research - 30 CV tailoring - 15 cover letters - 50 auto-apply - 10 interview prep - 5 salary negotiations - All job boards - Full analytics + export ## Plan comparison | Feature | Free | Starter | Pro | | --- | --- | --- | --- | | Jobs discovered/month | 50 | 500 | Unlimited | | AI scoring (Haiku) | 25 | 200 | 1,000 | | Deep research (Sonnet) | No | 20 | 100 | | CV tailoring (Sonnet) | No | 5 | 30 | | Cover letters (Opus) | No | 3 | 15 | | Auto-apply | No | 10/day | 50/day | | Interview prep | No | 3/mo | 10/mo | | Salary negotiation (Opus) | No | No | 5/mo | | Job boards | 3 boards | All boards | All boards | | Email outreach | No | Yes | Yes | | Salary explorer | 3/mo | 20/mo | Unlimited | | PPP comparison | No | 5 cities/mo | Unlimited | | Analytics | Basic | Full | Full + Export | | Career radar | View only | Full | Full | | A/B ML scoring | No | No | Yes | ## Billing questions ### Can I cancel anytime? Yes. Cancel from your settings page at any time. You keep access until the end of your billing period. ### What happens when I hit my limit? That feature is paused until your next billing period. Upgrade at any time to immediately restore access. ### Is there a free trial? Yes — Starter and Pro both come with a 7-day free trial. A payment method is required to start, but you won't be charged until the trial ends. ### What's the difference between weekly and monthly billing? Monthly billing includes roughly a 10% discount compared to paying weekly. Choose weekly if you're in an active job search and want flexibility; monthly if you're planning for the longer term. ### What payment methods do you accept? All major credit and debit cards (Visa, Mastercard, Amex) via Stripe. Apple Pay and Google Pay are also supported. ### Do you offer refunds? 7-day money-back guarantee on your first payment. After that, payments are non-refundable. Contact support if you have an issue. --- # Writing Canonical URL: https://quarrion.ai/blog > Long-form posts, each written to answer one question properly rather than to rank for a phrase. ## Can AI write your CV without making things up? Hallucinated skills are the documented failure mode of AI CV tools. Why it happens, why you often won't catch it, and the grounding architecture that prevents it. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/can-ai-write-your-cv-without-making-things-up Markdown: https://quarrion.ai/blog/can-ai-write-your-cv-without-making-things-up.md ## Comparing salaries across cities: the number on the offer is not the number A salary only means something after cost of living, inflation and currency are accounted for. How to compare offers across cities honestly, and where the data comes from. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/comparing-salaries-across-cities Markdown: https://quarrion.ai/blog/comparing-salaries-across-cities.md ## Following up with recruiters: timing, tone, and why almost nobody does it The follow-up is the cheapest high-leverage act in a job search and the most skipped. When to send one, what it should say, and what drafting automation fixes. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/following-up-with-recruiters Markdown: https://quarrion.ai/blog/following-up-with-recruiters.md ## Ghost jobs: why so many listings are dead, and how to stop applying into them A large share of job adverts are filled, paused or were never real. Where ghost listings come from, the signals that give them away, and how liveness checking works. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/ghost-jobs-and-dead-listings Markdown: https://quarrion.ai/blog/ghost-jobs-and-dead-listings.md ## How AI job matching actually works — and when a match score means anything Keyword overlap, embeddings and LLM judgement produce very different match scores. What each method can and cannot see, and how to tell which one you're getting. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/how-ai-job-matching-actually-works Markdown: https://quarrion.ai/blog/how-ai-job-matching-actually-works.md ## How many jobs should you apply to? Fewer than the internet says The applications-per-day question has the wrong unit. Why volume targets backfire, what actually limits a job search, and where the freed-up hours should go. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/how-many-jobs-should-you-apply-to Markdown: https://quarrion.ai/blog/how-many-jobs-should-you-apply-to.md ## How to track job applications without the spreadsheet dying by week three Why application-tracking spreadsheets always decay, what a working system has to record, and the point at which tracking is worth automating alongside the rest. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/how-to-track-job-applications Markdown: https://quarrion.ai/blog/how-to-track-job-applications.md ## What is Quarrion? An honest tour of the pipeline What the agent actually does at each stage — discovery, scoring, research, tailoring, applying and follow-up — and the parts it deliberately leaves to you. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/what-is-quarrion Markdown: https://quarrion.ai/blog/what-is-quarrion.md ## When an agent runs your job search, what's left for you to do? Delegating discovery, filtering and drafting frees most of a job search's hours. The highest-return places to reinvest them: skill gaps, interviews and your network. Published 2026-08-18 · 4 min read Read: https://quarrion.ai/blog/what-to-do-while-the-agent-searches Markdown: https://quarrion.ai/blog/what-to-do-while-the-agent-searches.md ## Do AI job application agents actually work? The honest answer depends entirely on which half of the problem the tool automates. Most automate the wrong half, and the difference shows up in your reply rate. Published 2026-08-10 · 5 min read Read: https://quarrion.ai/blog/do-ai-job-application-agents-work Markdown: https://quarrion.ai/blog/do-ai-job-application-agents-work.md ## Tailoring your CV per application: what it actually means Not rewriting it from scratch, and not stuffing it with keywords. A practical account of which parts of a CV should change per role, and which never should. Published 2026-08-08 · 4 min read Read: https://quarrion.ai/blog/tailoring-your-cv-per-application Markdown: https://quarrion.ai/blog/tailoring-your-cv-per-application.md ## Automating a UK job search without annoying employers Which parts of a UK job hunt are safe to automate, which are not, and the specifics that differ here — ATS platforms, right-to-work questions and salary bands. Published 2026-08-06 · 4 min read Read: https://quarrion.ai/blog/automating-a-uk-job-search Markdown: https://quarrion.ai/blog/automating-a-uk-job-search.md --- # Frequently Asked Questions Canonical URL: https://quarrion.ai/faq > How Quarrion discovers jobs, scores them against your CV, tailors each application, and handles your data. ### What does Quarrion actually do? Quarrion runs your job search as a continuous pipeline rather than a search box you have to keep revisiting. You give it your CV and a brief describing what you are looking for, and from then on the agent discovers roles across 100+ job boards, scores each one against your CV, tailors a CV and drafts a cover letter for the roles worth pursuing, completes application forms on your behalf, and drafts replies to recruiters for you to approve. You review and decide; it does the repetitive work. ### What do you need from me to get started? Your CV, and a short brief covering the roles you are targeting, the salary range you are looking for, and where you want to work. That is enough for the agent to start discovering and scoring roles. You can refine the brief at any time, and scoring reflects the change from the next run onwards. ### Do I need a credit card to try it? No. The Free plan requires no payment method at all — you can discover and AI-score jobs against your CV without entering card details. A card is only needed if you start a 7-day free trial of a paid plan, and you are not charged until that trial ends. ### Where do the jobs come from? The agent scans 100+ job boards. The Free plan covers a small subset of them; paid plans cover all of them. Discovery runs on a schedule in the background rather than only when you are signed in, so roles are found and scored while you are doing something else. ### Will I see the same job several times if it is posted to multiple boards? No. The same role advertised on several boards is de-duplicated before it reaches you, using both an exact content fingerprint and a fuzzy title-and-company match for the near-identical repostings that boards generate. So your monthly discovery allowance is spent on distinct roles, not on the same job counted five times. ### How does scoring work — is it just keyword matching? No. Each role is scored by a Claude model reading the actual content of your CV against the actual job description, so it can weigh relevant experience that shares no vocabulary with the posting — and, just as usefully, decline to reward a keyword you happen to have listed once. Keyword overlap is a poor proxy for fit in both directions, which is the reason the scoring stage exists at all. ### Which AI models do you use? Anthropic's Claude models, routed by task rather than one model for everything. Haiku handles the high-volume, repetitive work such as scoring and classification; Sonnet handles deep research and CV tailoring; Opus handles cover letters and salary negotiation. Routing the volume to the fast models is what makes it affordable to spend a genuinely strong model on the few outputs an employer will actually read. ### What is deep research? A slower, more thorough pass over a specific role and the company behind it, producing a research summary you can use when deciding whether to apply and when preparing for an interview. It is a paid-plan feature and is metered separately from scoring, because it costs considerably more to run per role. ### Will it apply to jobs for me? Yes — auto-apply completes and submits application forms on your behalf, on paid plans and within a daily cap. It is worth being clear about where responsibility sits: AI-generated CVs, cover letters and messages can contain mistakes, and reviewing what goes out before it goes out is your responsibility, not the agent's. That is set out in our Terms of Service, not just recommended here. ### Does it rewrite my CV for each role? On paid plans, yes. CV tailoring rewrites your CV against a specific job description, and a cover letter is drafted for that application. Each tailored document is produced for that one role, so your original CV stays intact and you can compare the two before sending anything. ### What happens after an application goes out? The agent tracks each application through its stages, drafts follow-up messages when a role goes quiet, and drafts replies to recruiter emails. Outbound messages are drafts for you to approve — the agent does not send correspondence to an employer on your behalf without you seeing it. Interview preparation is available on paid plans once a role reaches that stage. ### What happens to my CV and personal data? Your data is stored in Supabase-managed PostgreSQL. Your CV and job data may be sent to Anthropic's API to generate tailored documents, and payments are handled by Stripe — we never store your card details. You keep ownership of your CV content; using the service grants us only a limited licence to process it in order to provide the service. The full detail is on our Privacy Policy page. ### Can I delete my data? Yes. You can delete your account from the Settings page at any time, and your data is permanently removed within 30 days. Under GDPR you can also request a copy of your data, have it corrected, receive it in a portable format, restrict how it is processed, or object to processing — email the address published on our Privacy Policy page and we will action it. ### Do you track me across the web? No. We use essential cookies only, to keep you signed in and maintain your session. There are no third-party tracking cookies and no advertising cookies on this site. ### How much does it cost? There is a Free plan that stays free and needs no card, plus paid plans billed weekly or monthly in GBP through Stripe, each with a 7-day free trial. Current prices and the full per-plan allowances are on the pricing page — they are served there live from the billing system, which is why they are not repeated here where they could go stale. --- # Privacy Policy Canonical URL: https://quarrion.ai/privacy Last updated: March 2026 ## 1. Introduction Quarrion ("we", "our", "us") is committed to protecting your personal information. This Privacy Policy explains how we collect, use, and safeguard your data when you use our service. ## 2. Data We Collect - Account information: name, email address, password (hashed) - Job search preferences: target roles, salary range, locations - CV and cover letter content you upload or generate - Usage data: pages visited, features used, timestamps - Payment information: handled by Stripe (we do not store card details) ## 3. How We Use Your Data - To operate and improve the Quarrion pipeline - To generate AI-tailored CVs and cover letters using Anthropic's Claude models - To process payments via Stripe - To store your data securely in Supabase (PostgreSQL) - To send transactional emails (account confirmation, billing receipts) ## 4. Third-Party Services We use the following third-party services to operate our platform: - **Stripe**: payment processing. Stripe's Privacy Policy applies to payment data. - **Supabase**: database and authentication infrastructure. Your data is stored in Supabase-managed PostgreSQL. - **Anthropic**: AI language model provider (Claude). Your CV and job data may be processed by Anthropic's API to generate tailored documents. ## 5. Your Rights (GDPR) Under the General Data Protection Regulation (GDPR), you have the following rights: - **Right to Access**: request a copy of the personal data we hold about you. - **Right to Rectification**: request correction of inaccurate personal data. - **Right to Erasure**: request deletion of your personal data ("right to be forgotten"). - **Right to Portability**: receive your data in a structured, machine-readable format. - **Right to Object**: object to processing of your personal data for direct marketing purposes. - **Right to Restrict Processing**: request that we restrict how we use your data in certain circumstances. To exercise any of these rights, please contact us at natho1999@gmail.com. ## 6. Data Retention We retain your personal data for as long as your account is active. You may delete your account at any time from the Settings page. Upon deletion, your data will be permanently removed within 30 days. ## 7. Cookies We use essential cookies only — to maintain your session and authentication state. We do not use third-party tracking cookies or advertising cookies. ## 8. Contact Us For privacy inquiries, please email us at natho1999@gmail.com. --- # Terms of Service Canonical URL: https://quarrion.ai/terms Last updated: March 2026 ## 1. Acceptance of Terms By accessing or using Quarrion ("the Service"), you agree to be bound by these Terms of Service. If you do not agree, please do not use the Service. ## 2. Description of Service Quarrion is an AI-powered job search automation platform that discovers job opportunities, scores them against your profile, generates tailored CVs and cover letters, and manages application follow-ups on your behalf. ## 3. AI-Generated Content Disclaimer The Service uses AI language models (including Anthropic's Claude) to generate CVs, cover letters, outreach messages, and research summaries. AI-generated content may contain inaccuracies or errors. You are responsible for reviewing all AI-generated content before submitting it to employers. We make no warranty regarding the accuracy, completeness, or fitness for purpose of any AI-generated output. ## 4. Acceptable Use You agree not to use the Service to: - Send spam or unsolicited bulk messages to employers, recruiters, or other users - Submit false, misleading, or fraudulent job applications - Violate any applicable laws or regulations - Attempt to reverse-engineer, scrape, or otherwise misuse the platform - Use the Service on behalf of a third party without their consent We reserve the right to suspend or terminate your account if you violate these acceptable use requirements. ## 5. Subscription and Billing Paid plans are billed monthly or annually in advance. Prices are displayed in GBP. All payments are processed securely by Stripe. ## 6. Cancellation and Refunds You may cancel your subscription at any time from your account settings. Cancellation takes effect at the end of the current billing period — you retain access until then. We offer a 7-day money-back guarantee on your first payment. After that initial period, payments are non-refundable. If you believe you have been charged in error, please contact us at natho1999@gmail.com. ## 7. Intellectual Property You retain ownership of your CV content and personal data. By using the Service, you grant us a limited licence to process your content solely for the purpose of providing the Service. ## 8. Limitation of Liability To the maximum extent permitted by law, Quarrion shall not be liable for any indirect, incidental, or consequential damages arising from your use of the Service, including but not limited to lost job opportunities or employment outcomes. ## 9. Changes to Terms We may update these Terms from time to time. We will notify you by email or in-app notification before material changes take effect. Continued use of the Service constitutes acceptance of the updated Terms. ## 10. Contact For questions about these Terms, contact us at natho1999@gmail.com. --- # Developers Canonical URL: https://quarrion.ai/developers Quarrion publishes a small read-only API. It serves the same public data the website shows — plans and prices, the published FAQ, the advertised-salary medians behind the /salary pages, and the modelled estimates behind the salary globe — so an agent or a script can read it without scraping HTML. No key, no account, no signup. > This API is EXPERIMENTAL. It exists because publishing the data we already show is cheaper than being scraped, not because we are running an API product. Endpoints may change shape, move or disappear without notice. There is no versioning commitment and no uptime commitment. If you build something on it, email support@quarrion.ai so we know you are there — that is the only thing that will make us think twice before changing a field name. ## Endpoints Every response uses the envelope { "success": boolean, "data": …, "error": string | null }. Field casing follows whatever produces the payload: the plan and FAQ endpoints return database column names in snake_case, while the salary endpoints return camelCase values built by the aggregation service. ### GET /api/subscription/plans The live plan, price and per-metric limit table behind /pricing. Prices are in GBP. Owner-only rows are excluded. Served from a one-hour cache. No parameters. ``` curl 'https://quarrion.ai/api/subscription/plans' ``` ### GET /api/faq Every published entry, ordered by `sort_order` then newest first. Both filters are optional and combine; `search` is a case-insensitive substring match across the question and the answer. - `category` — Exact-match filter on the entry category. - `search` — Case-insensitive substring match on question or answer. ``` curl 'https://quarrion.ai/api/faq?category=product' ``` ### GET /api/salary/advertised The p25 / median / p75 of the annual pay employers STATED in public job advertisements, per sector or job title per market, over a rolling 90-day window. Every cell carries its own sample size, its date range, the nightly recompute run that produced it and the URL of its human-readable page. Advertised pay, not paid pay; an advertisement quoting no salary is excluded rather than estimated, and a cell is withdrawn rather than frozen when it falls below 30 advertisements. Covers every sector, not only technology. Both filters are optional and combine; `location` is an alias of `city` and `role` of `sector`, and either spelling — URL slug or raw value — resolves. Call /api/salary/advertised/locations first to learn the valid values. - `city` — Market: the URL slug (`london`, `uk`, `remote-uk`) or the raw location (`London`, `GB`, `remote-GB`). `location` is accepted as an alias. - `sector` — Sector or job title, as the URL slug (`hospitality-retail`, `chef-de-partie`) or its human form. `role` is accepted as an alias. ``` curl 'https://quarrion.ai/api/salary/advertised?city=london' ``` ### GET /api/salary/advertised/locations One entry per market, with how many cells it publishes, how many advertisements sit behind them, and the exact `city` / `sector` slugs to pass back to /api/salary/advertised. London leads because the corpus does; the rest follow by weight of evidence. No parameters. ``` curl 'https://quarrion.ai/api/salary/advertised/locations' ``` ### GET /api/salary/globe Country-level MODELLED ESTIMATES for a role, normalised to USD and inflation-adjusted. These are curated estimates calibrated against BLS / ONS / Eurostat reference data — not survey data, not observed pay, and not sourced from those agencies; every row carries a confidence score and its citations, and in production today every row resolves to `provenance.kind = "estimate"`. For advertised medians computed from real job postings, use /api/salary/advertised. Passing `country` switches the response to that country’s cities — the same payload as /api/salary/globe/cities. Anonymous callers read cross-tenant research data only; no user data is ever returned. - `role` (required) — Role title. Matched against a synonym list, so close variants resolve. - `country` — ISO 3166-1 alpha-2. When present the response is an array of cities instead. ``` curl 'https://quarrion.ai/api/salary/globe?role=Software%20Engineer' ``` ### GET /api/salary/globe/cities City-level drill-down of the same modelled estimates as /api/salary/globe — not observed pay. `median`, `p25` and `p75` are null for a city that is plotted but has no salary data for this role. - `role` (required) — Role title. Matched against a synonym list, so close variants resolve. - `country` (required) — ISO 3166-1 alpha-2 (alpha-3 is also accepted). ``` curl 'https://quarrion.ai/api/salary/globe/cities?role=Software%20Engineer&country=GB' ``` ### GET /api/salary/globe/roles Distinct role titles with researched salary data, so a client can suggest roles that will actually return results. Low-confidence free-text titles are filtered out. No parameters. ``` curl 'https://quarrion.ai/api/salary/globe/roles' ``` ### GET /api/health Liveness of the web tier and its database connection. Always answers 200 — a degraded database is reported in the body, not as a status code. This is the one public endpoint that does not send RateLimit headers. No parameters. ``` curl 'https://quarrion.ai/api/health' ``` ## Rate limits | Tier | Requests | Window | | --- | --- | --- | | free | 30 | 60s | | pro | 120 | 60s | | enterprise | 300 | 60s | Limits are per caller per 60 seconds. An anonymous caller is keyed by IP and gets the free-tier quota. Every response except /api/health carries RFC 9331 RateLimit and RateLimit-Policy headers. Those headers declare the policy in force rather than a live remaining count — a fabricated countdown would be worse than none, because you would pace against a number that means nothing. ## MCP server The same data is available over the Model Context Protocol, so an MCP client can call it as tools rather than as HTTP. The server is read-only: it exposes five tools, all of them wrappers over the endpoints above, and nothing that writes, authenticates or costs money. - Endpoint: https://quarrion.ai/api/v1/mcp - Discovery card: https://quarrion.ai/.well-known/mcp/server-card.json - `salary_advertised` — Advertised-salary medians from real job postings, with per-cell sample sizes. (wraps `GET /api/salary/advertised`) - `salary_globe` — Modelled country-level salary estimates for a role, with city drill-down. (wraps `GET /api/salary/globe`) - `salary_roles` — The role titles that have researched salary data. (wraps `GET /api/salary/globe/roles`) - `search_faq` — Search the published FAQ entries. (wraps `GET /api/faq`) - `get_pricing_plans` — The live plan, price and per-metric limit table. (wraps `GET /api/subscription/plans`) ## Machine-readable surfaces - https://quarrion.ai/openapi.json — OpenAPI 3.1 description of every endpoint on this page. - https://quarrion.ai/.well-known/api-catalog — RFC 9727 API catalog — a linkset pointing at the spec. - https://quarrion.ai/api/v1/mcp — Model Context Protocol endpoint (Streamable HTTP). - https://quarrion.ai/.well-known/mcp/server-card.json — MCP server discovery card. - https://quarrion.ai/.well-known/agent-skills/index.json — Agent-skills index. - https://quarrion.ai/llms.txt — Plain-text site summary with links to everything above. - https://quarrion.ai/llms-full.txt — Every public page as markdown, in one response. --- # Advertised Salaries Canonical URL: https://quarrion.ai/salary > Median advertised salary by role and sector, computed from the pay quoted in public job advertisements over a rolling 90-day window. These are ADVERTISED figures: the salary an employer published in a job advertisement. They are not a survey, not self-declared pay, and not a record of what anyone was actually paid. Base pay only — no bonus, tips, overtime or equity. Aggregated from public job listings on boards including Adzuna and Reed. Aggregate statistics only: no individual listing, employer name or advertisement text is published. Recomputed nightly; a figure is withdrawn if fewer than 30 postings support it. Data generation: run-2026-08-22T02:40:00.747Z. Per-cell detail, including the full method, is at each cell page with `.md` appended. Full methodology — how a figure is computed, what is excluded and what is never estimated: https://quarrion.ai/salary/methodology (markdown: https://quarrion.ai/salary/methodology.md) ## London ### Trades & Construction - Median advertised salary: £52,500 - 25th–75th percentile: £43,680–£66,799 - Based on 2335 London postings in the last 90 days (8 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/trades-construction - Markdown: https://quarrion.ai/salary/london/trades-construction.md ### Hospitality & Retail - Median advertised salary: £34,000 - 25th–75th percentile: £29,120–£40,000 - Based on 2242 London postings in the last 90 days (1 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/hospitality-retail - Markdown: https://quarrion.ai/salary/london/hospitality-retail.md ### Education - Median advertised salary: £37,000 - 25th–75th percentile: £27,950–£50,700 - Based on 2006 London postings in the last 90 days (11 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/education - Markdown: https://quarrion.ai/salary/london/education.md ### Operations & Supply Chain - Median advertised salary: £44,928 - 25th–75th percentile: £35,000–£55,864 - Based on 1657 London postings in the last 90 days (17 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/operations-supply-chain - Markdown: https://quarrion.ai/salary/london/operations-supply-chain.md ### Technology - Median advertised salary: £72,540 - 25th–75th percentile: £57,500–£91,000 - Based on 1558 London postings in the last 90 days (26 May 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/technology - Markdown: https://quarrion.ai/salary/london/technology.md ### Healthcare - Median advertised salary: £40,000 - 25th–75th percentile: £30,441–£52,403 - Based on 1258 London postings in the last 90 days (3 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/healthcare - Markdown: https://quarrion.ai/salary/london/healthcare.md ### Finance Accounting - Median advertised salary: £57,944 - 25th–75th percentile: £45,000–£75,088 - Based on 1084 London postings in the last 90 days (25 May 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/finance-accounting - Markdown: https://quarrion.ai/salary/london/finance-accounting.md ### Legal - Median advertised salary: £57,500 - 25th–75th percentile: £42,128–£75,280 - Based on 1025 London postings in the last 90 days (23 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/legal - Markdown: https://quarrion.ai/salary/london/legal.md ### Marketing Sales - Median advertised salary: £44,645 - 25th–75th percentile: £35,000–£55,619 - Based on 896 London postings in the last 90 days (27 May 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/marketing-sales - Markdown: https://quarrion.ai/salary/london/marketing-sales.md ### Consulting Professional Services - Median advertised salary: £61,670 - 25th–75th percentile: £50,000–£77,452 - Based on 658 London postings in the last 90 days (26 May 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/consulting-professional-services - Markdown: https://quarrion.ai/salary/london/consulting-professional-services.md ### HR People - Median advertised salary: £50,000 - 25th–75th percentile: £38,315–£63,155 - Based on 508 London postings in the last 90 days (12 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/hr-people - Markdown: https://quarrion.ai/salary/london/hr-people.md ### Government Public Sector - Median advertised salary: £56,763 - 25th–75th percentile: £45,000–£76,752 - Based on 340 London postings in the last 90 days (27 May 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/government-public-sector - Markdown: https://quarrion.ai/salary/london/government-public-sector.md ### Nonprofit - Median advertised salary: £36,167 - 25th–75th percentile: £31,097–£44,806 - Based on 316 London postings in the last 90 days (11 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/nonprofit - Markdown: https://quarrion.ai/salary/london/nonprofit.md ### Teaching Assistant - Median advertised salary: £26,650 - 25th–75th percentile: £25,415–£27,646 - Based on 213 London postings in the last 90 days (13 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/teaching-assistant - Markdown: https://quarrion.ai/salary/london/teaching-assistant.md ### Chef de Partie - Median advertised salary: £31,379 - 25th–75th percentile: £29,120–£34,985 - Based on 197 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/chef-de-partie - Markdown: https://quarrion.ai/salary/london/chef-de-partie.md ### Quantity Surveyor - Median advertised salary: £60,000 - 25th–75th percentile: £52,172–£65,516 - Based on 175 London postings in the last 90 days (8 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/quantity-surveyor - Markdown: https://quarrion.ai/salary/london/quantity-surveyor.md ### Senior Quantity Surveyor - Median advertised salary: £72,500 - 25th–75th percentile: £67,500–£80,000 - Based on 174 London postings in the last 90 days (1 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/senior-quantity-surveyor - Markdown: https://quarrion.ai/salary/london/senior-quantity-surveyor.md ### SEN Teaching Assistant - Median advertised salary: £26,650 - 25th–75th percentile: £25,350–£27,950 - Based on 154 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/sen-teaching-assistant - Markdown: https://quarrion.ai/salary/london/sen-teaching-assistant.md ### Creative Media - Median advertised salary: £41,649 - 25th–75th percentile: £35,129–£49,251 - Based on 128 London postings in the last 90 days (29 May 2026–16 August 2026). - Page: https://quarrion.ai/salary/london/creative-media - Markdown: https://quarrion.ai/salary/london/creative-media.md ### Electrician - Median advertised salary: £48,464 - 25th–75th percentile: £44,000–£57,782 - Based on 121 London postings in the last 90 days (2 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/electrician - Markdown: https://quarrion.ai/salary/london/electrician.md ### Project Manager - Median advertised salary: £63,019 - 25th–75th percentile: £55,000–£85,000 - Based on 120 London postings in the last 90 days (1 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/project-manager - Markdown: https://quarrion.ai/salary/london/project-manager.md ### Sous Chef - Median advertised salary: £40,000 - 25th–75th percentile: £35,000–£44,633 - Based on 117 London postings in the last 90 days (2 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/sous-chef - Markdown: https://quarrion.ai/salary/london/sous-chef.md ### Science Research - Median advertised salary: £47,523 - 25th–75th percentile: £43,299–£55,000 - Based on 109 London postings in the last 90 days (18 June 2026–19 August 2026). - Page: https://quarrion.ai/salary/london/science-research - Markdown: https://quarrion.ai/salary/london/science-research.md ### Support Worker - Median advertised salary: £28,333 - 25th–75th percentile: £27,268–£29,863 - Based on 104 London postings in the last 90 days (26 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/support-worker - Markdown: https://quarrion.ai/salary/london/support-worker.md ### Head Chef - Median advertised salary: £46,148 - 25th–75th percentile: £43,016–£53,400 - Based on 95 London postings in the last 90 days (2 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/head-chef - Markdown: https://quarrion.ai/salary/london/head-chef.md ### Site Manager - Median advertised salary: £62,500 - 25th–75th percentile: £55,000–£70,150 - Based on 94 London postings in the last 90 days (1 June 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/site-manager - Markdown: https://quarrion.ai/salary/london/site-manager.md ### Assistant Manager - Median advertised salary: £35,000 - 25th–75th percentile: £31,614–£39,036 - Based on 87 London postings in the last 90 days (8 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/assistant-manager - Markdown: https://quarrion.ai/salary/london/assistant-manager.md ### Graduate Teaching Assistant - Median advertised salary: £26,650 - 25th–75th percentile: £26,390–£27,500 - Based on 81 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/graduate-teaching-assistant - Markdown: https://quarrion.ai/salary/london/graduate-teaching-assistant.md ### Sales Assistant - Median advertised salary: £25,376 - 25th–75th percentile: £24,024–£27,955 - Based on 81 London postings in the last 90 days (26 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/sales-assistant - Markdown: https://quarrion.ai/salary/london/sales-assistant.md ### Plumber - Median advertised salary: £40,000 - 25th–75th percentile: £38,000–£49,400 - Based on 75 London postings in the last 90 days (12 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/plumber - Markdown: https://quarrion.ai/salary/london/plumber.md ### Care Assistant - Median advertised salary: £30,784 - 25th–75th percentile: £28,465–£33,737 - Based on 73 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/care-assistant - Markdown: https://quarrion.ai/salary/london/care-assistant.md ### Learning Support Assistant - Median advertised salary: £27,950 - 25th–75th percentile: £26,650–£28,600 - Based on 73 London postings in the last 90 days (15 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/learning-support-assistant - Markdown: https://quarrion.ai/salary/london/learning-support-assistant.md ### Store Manager - Median advertised salary: £40,000 - 25th–75th percentile: £35,000–£47,238 - Based on 67 London postings in the last 90 days (9 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/store-manager - Markdown: https://quarrion.ai/salary/london/store-manager.md ### Electrical Maintenance Engineer - Median advertised salary: £45,000 - 25th–75th percentile: £45,000–£48,900 - Based on 65 London postings in the last 90 days (4 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/electrical-maintenance-engineer - Markdown: https://quarrion.ai/salary/london/electrical-maintenance-engineer.md ### Restaurant Manager - Median advertised salary: £43,249 - 25th–75th percentile: £38,639–£46,500 - Based on 65 London postings in the last 90 days (24 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/restaurant-manager - Markdown: https://quarrion.ai/salary/london/restaurant-manager.md ### Cover Supervisor - Median advertised salary: £27,950 - 25th–75th percentile: £26,910–£28,600 - Based on 62 London postings in the last 90 days (27 July 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/cover-supervisor - Markdown: https://quarrion.ai/salary/london/cover-supervisor.md ### SEN Teacher - Median advertised salary: £50,850 - 25th–75th percentile: £47,304–£59,150 - Based on 62 London postings in the last 90 days (24 July 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/sen-teacher - Markdown: https://quarrion.ai/salary/london/sen-teacher.md ### Primary Teacher - Median advertised salary: £51,000 - 25th–75th percentile: £46,800–£61,230 - Based on 59 London postings in the last 90 days (5 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/primary-teacher - Markdown: https://quarrion.ai/salary/london/primary-teacher.md ### Data Engineer - Median advertised salary: £62,433 - 25th–75th percentile: £55,553–£75,000 - Based on 58 London postings in the last 90 days (30 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/data-engineer - Markdown: https://quarrion.ai/salary/london/data-engineer.md ### General Manager - Median advertised salary: £50,000 - 25th–75th percentile: £43,372–£60,000 - Based on 58 London postings in the last 90 days (16 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/general-manager - Markdown: https://quarrion.ai/salary/london/general-manager.md ### Operations Manager - Median advertised salary: £52,500 - 25th–75th percentile: £43,727–£65,000 - Based on 56 London postings in the last 90 days (28 May 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/operations-manager - Markdown: https://quarrion.ai/salary/london/operations-manager.md ### Software Engineer - Median advertised salary: £80,000 - 25th–75th percentile: £75,000–£95,000 - Based on 55 London postings in the last 90 days (2 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/software-engineer - Markdown: https://quarrion.ai/salary/london/software-engineer.md ### Assistant Quantity Surveyor - Median advertised salary: £40,000 - 25th–75th percentile: £38,357–£45,000 - Based on 54 London postings in the last 90 days (5 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/assistant-quantity-surveyor - Markdown: https://quarrion.ai/salary/london/assistant-quantity-surveyor.md ### Functional Assessor - Median advertised salary: £45,000 - 25th–75th percentile: £37,186–£45,000 - Based on 54 London postings in the last 90 days (18 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/functional-assessor - Markdown: https://quarrion.ai/salary/london/functional-assessor.md ### Data Analyst - Median advertised salary: £50,551 - 25th–75th percentile: £46,587–£58,779 - Based on 54 London postings in the last 90 days (25 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/data-analyst - Markdown: https://quarrion.ai/salary/london/data-analyst.md ### Kitchen Porter - Median advertised salary: £27,259 - 25th–75th percentile: £25,424–£28,927 - Based on 54 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/kitchen-porter - Markdown: https://quarrion.ai/salary/london/kitchen-porter.md ### Assistant Store Manager - Median advertised salary: £30,381 - 25th–75th percentile: £28,422–£35,180 - Based on 47 London postings in the last 90 days (7 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/assistant-store-manager - Markdown: https://quarrion.ai/salary/london/assistant-store-manager.md ### Chef - Median advertised salary: £29,120 - 25th–75th percentile: £26,758–£33,280 - Based on 45 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/chef - Markdown: https://quarrion.ai/salary/london/chef.md ### Legal Counsel - Median advertised salary: £92,407 - 25th–75th percentile: £73,033–£100,000 - Based on 44 London postings in the last 90 days (10 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/legal-counsel - Markdown: https://quarrion.ai/salary/london/legal-counsel.md ### Executive Assistant - Median advertised salary: £47,961 - 25th–75th percentile: £42,250–£58,750 - Based on 43 London postings in the last 90 days (30 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/executive-assistant - Markdown: https://quarrion.ai/salary/london/executive-assistant.md ### Assistant Restaurant Manager - Median advertised salary: £38,800 - 25th–75th percentile: £34,410–£45,000 - Based on 42 London postings in the last 90 days (24 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/assistant-restaurant-manager - Markdown: https://quarrion.ai/salary/london/assistant-restaurant-manager.md ### Senior Project Manager - Median advertised salary: £71,497 - 25th–75th percentile: £65,000–£81,415 - Based on 42 London postings in the last 90 days (5 June 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/senior-project-manager - Markdown: https://quarrion.ai/salary/london/senior-project-manager.md ### Send Teaching Assistant - Median advertised salary: £26,000 - 25th–75th percentile: £25,350–£26,650 - Based on 41 London postings in the last 90 days (24 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/send-teaching-assistant - Markdown: https://quarrion.ai/salary/london/send-teaching-assistant.md ### Business Development Manager - Median advertised salary: £51,186 - 25th–75th percentile: £45,873–£63,125 - Based on 40 London postings in the last 90 days (3 June 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/business-development-manager - Markdown: https://quarrion.ai/salary/london/business-development-manager.md ### HR Business Partner - Median advertised salary: £67,500 - 25th–75th percentile: £54,875–£75,625 - Based on 40 London postings in the last 90 days (3 June 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/hr-business-partner - Markdown: https://quarrion.ai/salary/london/hr-business-partner.md ### Data Architect - Median advertised salary: £80,000 - 25th–75th percentile: £65,395–£85,522 - Based on 40 London postings in the last 90 days (16 June 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/data-architect - Markdown: https://quarrion.ai/salary/london/data-architect.md ### Supervisor - Median advertised salary: £29,469 - 25th–75th percentile: £27,489–£31,559 - Based on 40 London postings in the last 90 days (8 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/supervisor - Markdown: https://quarrion.ai/salary/london/supervisor.md ### Finance Manager - Median advertised salary: £62,500 - 25th–75th percentile: £56,250–£72,500 - Based on 39 London postings in the last 90 days (13 July 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/finance-manager - Markdown: https://quarrion.ai/salary/london/finance-manager.md ### Recruitment Consultant - Median advertised salary: £34,000 - 25th–75th percentile: £29,250–£39,389 - Based on 39 London postings in the last 90 days (4 June 2026–19 August 2026). - Page: https://quarrion.ai/salary/london/recruitment-consultant - Markdown: https://quarrion.ai/salary/london/recruitment-consultant.md ### Senior Software Engineer - Median advertised salary: £82,500 - 25th–75th percentile: £68,091–£90,000 - Based on 38 London postings in the last 90 days (27 May 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/senior-software-engineer - Markdown: https://quarrion.ai/salary/london/senior-software-engineer.md ### Senior Data Analyst - Median advertised salary: £52,206 - 25th–75th percentile: £48,178–£65,500 - Based on 36 London postings in the last 90 days (17 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/senior-data-analyst - Markdown: https://quarrion.ai/salary/london/senior-data-analyst.md ### Senior Legal Counsel - Median advertised salary: £83,797 - 25th–75th percentile: £75,000–£100,000 - Based on 36 London postings in the last 90 days (1 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/senior-legal-counsel - Markdown: https://quarrion.ai/salary/london/senior-legal-counsel.md ### Engineering Manager - Median advertised salary: £78,479 - 25th–75th percentile: £69,084–£90,000 - Based on 35 London postings in the last 90 days (20 July 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/engineering-manager - Markdown: https://quarrion.ai/salary/london/engineering-manager.md ### Hotel Kitchen Team Member - Median advertised salary: £31,480 - 25th–75th percentile: £29,740–£33,803 - Based on 34 London postings in the last 90 days (30 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/hotel-kitchen-team-member - Markdown: https://quarrion.ai/salary/london/hotel-kitchen-team-member.md ### Supply Teacher - Median advertised salary: £40,950 - 25th–75th percentile: £37,570–£43,550 - Based on 34 London postings in the last 90 days (22 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/supply-teacher - Markdown: https://quarrion.ai/salary/london/supply-teacher.md ### Team Member - Median advertised salary: £25,386 - 25th–75th percentile: £24,614–£28,600 - Based on 34 London postings in the last 90 days (28 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/team-member - Markdown: https://quarrion.ai/salary/london/team-member.md ### Higher Level Teaching Assistant - Median advertised salary: £33,150 - 25th–75th percentile: £28,600–£38,350 - Based on 33 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/higher-level-teaching-assistant - Markdown: https://quarrion.ai/salary/london/higher-level-teaching-assistant.md ### Hotel Duty Manager - Median advertised salary: £35,795 - 25th–75th percentile: £35,075–£38,442 - Based on 33 London postings in the last 90 days (25 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/hotel-duty-manager - Markdown: https://quarrion.ai/salary/london/hotel-duty-manager.md ### Paralegal - Median advertised salary: £34,000 - 25th–75th percentile: £29,500–£36,794 - Based on 33 London postings in the last 90 days (17 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/paralegal - Markdown: https://quarrion.ai/salary/london/paralegal.md ### Employment Solicitor - Median advertised salary: £71,356 - 25th–75th percentile: £62,291–£87,500 - Based on 32 London postings in the last 90 days (15 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/employment-solicitor - Markdown: https://quarrion.ai/salary/london/employment-solicitor.md ### Multi Trader - Median advertised salary: £45,146 - 25th–75th percentile: £39,248–£48,503 - Based on 32 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/multi-trader - Markdown: https://quarrion.ai/salary/london/multi-trader.md ### Mechanical Project Manager - Median advertised salary: £82,500 - 25th–75th percentile: £77,583–£90,000 - Based on 32 London postings in the last 90 days (18 June 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/mechanical-project-manager - Markdown: https://quarrion.ai/salary/london/mechanical-project-manager.md ### Project Quantity Surveyor - Median advertised salary: £58,125 - 25th–75th percentile: £55,000–£62,500 - Based on 32 London postings in the last 90 days (16 June 2026–19 August 2026). - Page: https://quarrion.ai/salary/london/project-quantity-surveyor - Markdown: https://quarrion.ai/salary/london/project-quantity-surveyor.md ### Registered Nurse - Median advertised salary: £47,388 - 25th–75th percentile: £36,911–£52,000 - Based on 32 London postings in the last 90 days (16 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/registered-nurse - Markdown: https://quarrion.ai/salary/london/registered-nurse.md ### Senior Data Engineer - Median advertised salary: £78,750 - 25th–75th percentile: £67,500–£95,000 - Based on 32 London postings in the last 90 days (18 June 2026–22 August 2026). - Page: https://quarrion.ai/salary/london/senior-data-engineer - Markdown: https://quarrion.ai/salary/london/senior-data-engineer.md ### Account Manager - Median advertised salary: £50,000 - 25th–75th percentile: £42,256–£55,862 - Based on 31 London postings in the last 90 days (24 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/account-manager - Markdown: https://quarrion.ai/salary/london/account-manager.md ### Branch Manager - Median advertised salary: £42,074 - 25th–75th percentile: £39,999–£52,466 - Based on 31 London postings in the last 90 days (27 July 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/branch-manager - Markdown: https://quarrion.ai/salary/london/branch-manager.md ### Commis Chef - Median advertised salary: £29,636 - 25th–75th percentile: £27,820–£32,361 - Based on 31 London postings in the last 90 days (27 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/commis-chef - Markdown: https://quarrion.ai/salary/london/commis-chef.md ### Contracts Manager - Median advertised salary: £51,447 - 25th–75th percentile: £50,000–£83,750 - Based on 31 London postings in the last 90 days (24 July 2026–20 August 2026). - Page: https://quarrion.ai/salary/london/contracts-manager - Markdown: https://quarrion.ai/salary/london/contracts-manager.md ### Healthcare Assistant - Median advertised salary: £28,731 - 25th–75th percentile: £26,725–£30,784 - Based on 30 London postings in the last 90 days (17 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/healthcare-assistant - Markdown: https://quarrion.ai/salary/london/healthcare-assistant.md ### Junior Sous Chef - Median advertised salary: £37,653 - 25th–75th percentile: £34,196–£40,044 - Based on 30 London postings in the last 90 days (28 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/junior-sous-chef - Markdown: https://quarrion.ai/salary/london/junior-sous-chef.md ### HR Manager - Median advertised salary: £62,500 - 25th–75th percentile: £47,516–£70,000 - Based on 30 London postings in the last 90 days (17 June 2026–19 August 2026). - Page: https://quarrion.ai/salary/london/hr-manager - Markdown: https://quarrion.ai/salary/london/hr-manager.md ### Senior Engineer - Median advertised salary: £78,214 - 25th–75th percentile: £71,138–£100,000 - Based on 30 London postings in the last 90 days (28 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/london/senior-engineer - Markdown: https://quarrion.ai/salary/london/senior-engineer.md ## United Kingdom ### Trades & Construction - Median advertised salary: £70,000 - 25th–75th percentile: £50,000–£95,000 - Based on 185 UK postings in the last 90 days (26 May 2026–19 August 2026). - Page: https://quarrion.ai/salary/uk/trades-construction - Markdown: https://quarrion.ai/salary/uk/trades-construction.md ### Operations & Supply Chain - Median advertised salary: £60,000 - 25th–75th percentile: £45,000–£169,320 - Based on 73 UK postings in the last 90 days (26 May 2026–13 August 2026). - Page: https://quarrion.ai/salary/uk/operations-supply-chain - Markdown: https://quarrion.ai/salary/uk/operations-supply-chain.md ### Technology - Median advertised salary: £80,000 - 25th–75th percentile: £51,500–£133,431 - Based on 70 UK postings in the last 90 days (26 May 2026–7 August 2026). - Page: https://quarrion.ai/salary/uk/technology - Markdown: https://quarrion.ai/salary/uk/technology.md ### Healthcare - Median advertised salary: £63,788 - 25th–75th percentile: £33,236–£115,159 - Based on 46 UK postings in the last 90 days (26 May 2026–29 July 2026). - Page: https://quarrion.ai/salary/uk/healthcare - Markdown: https://quarrion.ai/salary/uk/healthcare.md ## Remote (US) ### Technology - Median advertised salary: $157,500 - 25th–75th percentile: $125,000–$187,500 - Based on 313 remote US postings in the last 90 days (29 May 2026–21 August 2026). - Page: https://quarrion.ai/salary/remote-us/technology - Markdown: https://quarrion.ai/salary/remote-us/technology.md ### Consulting Professional Services - Median advertised salary: $179,900 - 25th–75th percentile: $160,000–$197,500 - Based on 45 remote US postings in the last 90 days (5 June 2026–8 August 2026). - Page: https://quarrion.ai/salary/remote-us/consulting-professional-services - Markdown: https://quarrion.ai/salary/remote-us/consulting-professional-services.md ## San Francisco ### Technology - Median advertised salary: $200,000 - 25th–75th percentile: $190,000–$230,000 - Based on 149 San Francisco postings in the last 90 days (30 July 2026–21 August 2026). - Page: https://quarrion.ai/salary/san-francisco/technology - Markdown: https://quarrion.ai/salary/san-francisco/technology.md ## New York ### Technology - Median advertised salary: $170,000 - 25th–75th percentile: $146,222–$205,000 - Based on 58 New York postings in the last 90 days (30 July 2026–20 August 2026). - Page: https://quarrion.ai/salary/new-york/technology - Markdown: https://quarrion.ai/salary/new-york/technology.md ## Remote (UK) ### Technology - Median advertised salary: £110,000 - 25th–75th percentile: £90,000–£131,523 - Based on 50 remote UK postings in the last 90 days (29 June 2026–21 August 2026). - Page: https://quarrion.ai/salary/remote-uk/technology - Markdown: https://quarrion.ai/salary/remote-uk/technology.md ## Manchester ### Technology - Median advertised salary: £57,500 - 25th–75th percentile: £48,500–£70,000 - Based on 37 Manchester postings in the last 90 days (3 June 2026–13 August 2026). - Page: https://quarrion.ai/salary/manchester/technology - Markdown: https://quarrion.ai/salary/manchester/technology.md --- # How we compute advertised salaries Canonical URL: https://quarrion.ai/salary/methodology > Every published figure is a median of the annual pay employers stated in public job advertisements over a rolling 90-day window, recomputed nightly, published only where at least 30 advertisements support it. This page states the method in full: what a single observation is, which advertisements are excluded and why, what gets a whole figure withdrawn, and where the numbers come from. It is written to be checked rather than taken on trust — every threshold named below is the one the code enforces. ## What these figures are They are ADVERTISED salaries: the annual pay an employer published in a job advertisement. That is the whole claim, and it is deliberately a narrow one. - **Not a survey**: Nobody was asked what they earn. No figure here is self-declared by an employee, and none is weighted to a population. - **Not what anyone was paid**: An advertised range is an opening position. An employer may settle above or below the figure it advertised, and the settlement is not public. - **Base pay only**: Bonus, tips, overtime, commission, equity and benefits are outside what an advertisement usually quotes, so they are outside these figures. - **Aggregates only**: No individual listing, employer name or line of advertisement text is published anywhere on this site. A cell is a count and three percentiles. ## How the figures are computed A cell is one role in one market — "chef de partie in London". Every advertisement matching that cell and carrying an annual salary in the window becomes one observation, and the cell is the 25th percentile, the median and the 75th percentile of that sample. - **One observation per advertisement**: The midpoint of the advertised range when both bounds are given, or the single bound when only one is. A missing bound is treated as absent, never as zero — averaging in a zero is how a median gets dragged toward nothing. - **A rolling 90-day window**: Measured on the date the advertisement was posted. A posting date in the future is refused rather than accepted, because a window that ends after today is visibly impossible and discredits every number beside it. - **Converted before aggregating**: Each cell is priced in its own local market currency, and every observation is converted into that currency before it joins the sample. An advertisement in a currency we cannot convert is dropped, never averaged in as though it were already local. - **A plausibility band, applied after conversion**: An annual figure below 15,000 or above 500,000 in the cell's currency is not a salary and is dropped. After conversion, because the same raw number can be nonsense in one currency and ordinary in another. - **A floor of 30 advertisements**: A cell below the floor is not published at all, and a published cell that falls below it is WITHDRAWN rather than frozen at a number we can no longer stand behind. That is why a page you bookmarked can legitimately disappear. - **Outlier suppression on the spread**: A cell whose 75th percentile exceeds 4 times its 25th is suppressed entirely. A spread that wide is the signature of two different jobs sharing a title, and a median across them describes neither. - **Recomputed nightly**: The whole table is rebuilt each night from the current window, so a figure is never carried forward. Every cell states when it was last computed. ## What is excluded, and never estimated An advertisement with no stated annual salary is EXCLUDED. It is not imputed, modelled or predicted from the job title, and no estimated figure is ever mixed into a published median. If a cell has too few advertisements that stated pay, the cell does not publish — the gap is left visible rather than filled in. - **No stated annual salary**: Excluded. "Competitive" and "DOE" contribute nothing to any figure here. - **Hourly, daily, weekly and monthly rates**: Excluded rather than annualised. Annualising a day rate requires assuming the working pattern, and the assumption would silently become the answer. - **Unconvertible currencies**: Excluded and counted, never defaulted to a currency we merely guessed at. - **Advertisement copy dressed as a job title**: Titles longer than 40 characters, or containing pipes, brackets, slashes or digits, do not form a cell — those are reference codes, pay fragments and marketing lines, not occupations. The honest consequence, stated rather than buried: a median over advertisements that stated pay is not a median over all advertisements. Where employers who publish a salary differ systematically from those who do not, these figures inherit that difference. Excluding is still the better trade — a stated number can be wrong, but an estimated one cannot even be checked. ## Where the data comes from Aggregated from public job listings on boards including Adzuna and Reed. Coverage is whatever those boards carry, which is why the published cells span trades and construction, hospitality and retail, education, healthcare, care work, logistics and office roles rather than one industry. The same vacancy re-posted by three agencies counts as three advertised observations. The corpus already de-duplicates on content at ingest; suppressing more than that would require employer-level identity we do not publish, and guessing at it would remove real adverts as often as duplicates. ## Freshness and provenance - **Last computed**: Every cell page prints the date its figures were computed. The recompute runs nightly, so this is normally yesterday or today. - **Data generation**: Each recompute stamps every row it writes with one generation id, printed on every salary surface — this page's siblings, /salary, each cell, each markdown twin and /llms.txt. Two of our surfaces quoting different numbers with different generation ids are two snapshots, not a contradiction; the same generation id with different numbers would be a bug worth reporting. - **Window dates**: Every cell states the first and last posting date in its own sample, so the window is checkable rather than assumed from this page. ## How this differs from other salary sources Advertised-salary data is well-established and several organisations publish it. These are the choices this dataset makes, so you can judge whether they suit the question you are asking — not a claim about anyone else. - **Every sector the boards carry**: Chefs, teaching assistants, care workers, electricians and quantity surveyors are cells here on the same terms as software engineers. Nothing is scoped to one industry. - **Stated pay only**: Advertisements without a salary are excluded, never estimated, so no published figure is partly a model output. - **A median, not a mean**: A mean advertised salary moves with a handful of outlying adverts. The quartiles are published alongside it so the spread is visible. - **The sample is on the figure**: The posting count and the date range sit beside every median, on the index, on the page and in the machine-readable copies — not in a footnote. - **Readable by machine**: Every page here has a markdown twin, the whole table is available as one document, and the same figures are served over the public API. Every published figure: https://quarrion.ai/salary (markdown: https://quarrion.ai/salary.md) --- # Can AI write your CV without making things up? Canonical URL: https://quarrion.ai/blog/can-ai-write-your-cv-without-making-things-up > Hallucinated skills are the documented failure mode of AI CV tools. Why it happens, why you often won't catch it, and the grounding architecture that prevents it. Published 2026-08-18 · 4 min read Ask a general-purpose chatbot to "improve" your CV against a job posting and read the output carefully — genuinely carefully, line by line. There's a real chance it now says something you never did. A technology you've never touched, promoted from the posting's requirements into your experience. An accomplishment quietly inflated to rounder numbers. A responsibility that belonged to your team, reassigned to you personally. This isn't a horror story about one bad tool. It's the default behaviour of the underlying technology, and understanding why is the difference between using AI on your CV safely and discovering the problem in an interview. ## Why models invent experience Language models are trained to produce plausible text, and their notion of plausible comes from the shape of the documents they trained on. A strong CV for a data-engineering role *usually* mentions certain tools. Given your CV, the posting, and the instruction "make this fit", the statistically likely completion includes those tools — whether or not they were in the input. The model isn't lying, exactly; it has no model of which claims about you are *true*, only of which claims are *typical*. When instructed to close the gap between your CV and a posting, fabrication is the shortest path, and nothing in the objective penalises it. The failure is insidious because of *where* it lands. The invented material is, by construction, exactly what the posting asked for — so it reads as perfectly natural in context. You wrote the true 95 percent of the document, you're skimming your own familiar prose, and the seams don't show. The claims most likely to be fabricated are precisely the ones a reviewer is scanning for, and precisely the ones an interviewer will probe. ## What it costs when it surfaces An invented skill doesn't fail at the CV screen — it passes the CV screen. That's the problem. It fails twenty minutes into a technical interview, when someone asks a concrete question about the experience you supposedly have, in front of exactly the audience you most wanted to impress. The interviewer can't tell an AI fabrication from a candidate's lie, and the charitable interpretation isn't available to them: the CV is your document, sent under your name. In regulated fields the stakes go up from embarrassing to disqualifying. And the reputational ledger is asymmetric — one caught fabrication colours every true claim on the page. ## Grounding: the fix that's architectural, not behavioural The unreliable fix is prompting the model to be honest — "don't invent anything" reduces fabrication without eliminating it, because the instruction is fighting the training objective. The reliable fix changes what the model is allowed to draw from. Quarrion's tailoring is grounded in a structured profile built from your actual history — a knowledge graph of your roles, projects, skills and accomplishments, extracted from your real CV and the information you've provided. Tailoring for a posting then works by **selection and emphasis**: choosing which of *your* verified experiences to foreground, reordering, reframing vocabulary toward the posting's, trimming what's irrelevant. The generator can't claim Kubernetes experience unless Kubernetes exists in your graph, because the graph is the only place claims are allowed to come from. Fabrication isn't discouraged; it's unrepresentable. That's the distinction worth taking to any tool in this category: does it generate from your verified history, or generate plausible text and hope? The first architecture makes honesty structural. The second makes it a behaviour you have to audit for, forever. ## What honest tailoring still does Ruling out invention doesn't reduce tailoring to formatting. Legitimate, grounded tailoring still moves the needle: surfacing the two genuinely relevant projects that were buried under chronology; translating your vocabulary to the posting's where they name the same skill; cutting the half-page that means nothing for this role; leading with the achievement this hiring manager will care about. All of it is your real experience, arranged for this reader — which is what tailoring was supposed to mean before the shortcut became available. The gap between your experience and a posting's wishlist, where it's real, is information. Sometimes it says "apply anyway, and be ready to talk about the adjacent thing you did instead" — a case honest materials let you make credibly. Sometimes it says this role isn't the one, and the evening belongs to a better-matched application. A tool that papers over the gap destroys the signal both ways. ## The test worth running Before trusting any AI CV tool with something sent under your name: feed it a posting requiring a skill you demonstrably lack, and see what comes back. A grounded system works with what you have or leaves the gap visible. A plausibility engine gives you the skill. It's a two-minute experiment, and it tells you which kind of ghostwriter you've hired — one constrained by your record, or one constrained by nothing. --- # Comparing salaries across cities: the number on the offer is not the number Canonical URL: https://quarrion.ai/blog/comparing-salaries-across-cities > A salary only means something after cost of living, inflation and currency are accounted for. How to compare offers across cities honestly, and where the data comes from. Published 2026-08-18 · 4 min read Is £68,000 in London a better offer than €72,000 in Berlin? The honest answer is that the question, as posed, is unanswerable — not because it's hard but because the two numbers aren't in the same units. They differ by currency, by what a unit of currency buys locally, and — if you're comparing against a benchmark from even a couple of years ago — by when the number was true. A useful comparison converts everything into the same units first. Most salary conversations never do, which is why they generate more heat than information. ## The three conversions that matter **Currency is the shallow one.** Exchange rates move constantly, and a comparison made at last year's rate can be wrong by a meaningful margin today. Any tool doing cross-border comparison needs live FX, not a rate baked in when the page was built. But currency conversion alone answers the wrong question — it tells you what the Berlin salary would buy *in London*, which is only relevant if you plan to earn in Berlin and spend in London. **Cost of living is the deep one.** What you actually want to compare is the life each salary buys where it's earned: rent, groceries, transport, childcare. Rent alone can differ between cities by a factor that swallows the entire gap between two offers. Cost-of-living indices — Numbeo's crowdsourced data is the widely used source — let you deflate each salary by local prices, turning "what you're paid" into "what you can live like". On that measure, the lower nominal offer wins surprisingly often; expensive cities are expensive precisely because the salary premium rarely covers the price premium. **Inflation is the one everyone forgets.** Salary data ages badly, and it aged very badly through the recent high-inflation years. A benchmark collected three years ago understates today's market by the accumulated price growth since — so a role that "pays above the median" against a stale benchmark may pay below the real one. Honest benchmarks state their collection date and adjust via CPI to present-day terms. Benchmarks that don't mention a date are stale until proven otherwise. There's a fourth lens for international moves: purchasing power parity, which compares currencies by what they buy rather than what they trade at. PPP is the right frame for "would my standard of living rise?", while market FX is the right frame for anything you'll spend abroad — remittances, foreign savings, travel. The two frames can point in opposite directions for the same offer, and knowing which question you're asking is most of the work. ## Doing it by hand The manual version of this comparison: find a benchmark for the role in each city and check its date. Adjust old benchmarks forward by inflation. Convert currencies at today's rate for the nominal view. Deflate each salary by a cost-of-living index for the real view. Then look at both views, because they answer different questions — the nominal view is what your savings rate looks like if you live frugally; the real view is what daily life feels like if you don't. It's an hour of spreadsheet work per comparison, done carefully. The failure mode isn't inability — every ingredient is public — it's that nobody re-does an hour of spreadsheet work every time an offer moves or a new city enters the conversation, so decisions get made on the last comparison, not the current one. ## What Quarrion does with this Quarrion keeps a salary layer with city-level benchmarks that have the corrections applied: inflation-adjusted to present-day terms, comparable across cities via cost-of-living data, convertible at live exchange rates, with a PPP view for cross-country questions — browsable on an interactive globe, because the geography is genuinely part of the information. The point isn't decoration; it's that "is this offer good?" becomes a question you can answer during a negotiation rather than after it. That timing matters more than it looks. The moment you most need a defensible number is the moment a recruiter asks your expectations — early, on the phone, anchoring everything after. Arriving with a current, adjusted, city-specific range changes that conversation from guessing to negotiating. ## The caveats that keep it honest Benchmarks are distributions, not promises — your offer is set by a company's band and your negotiation, not by a median. Crowdsourced cost-of-living data runs on contributions, and thinly covered cities carry wider error bars. Taxes, healthcare and pension differences between countries can move a comparison as much as rent does, and no index fully captures them. And some things worth moving for — proximity to family, a language you want to live in, a city you love — never appear in any dataset. The comparison won't make the decision for you. What it does is smaller and still valuable: it stops the decision being made by a number that was never in the right units to begin with. --- # Following up with recruiters: timing, tone, and why almost nobody does it Canonical URL: https://quarrion.ai/blog/following-up-with-recruiters > The follow-up is the cheapest high-leverage act in a job search and the most skipped. When to send one, what it should say, and what drafting automation fixes. Published 2026-08-18 · 4 min read Ask recruiters what candidates get wrong and the answers cluster around two poles: the candidates who never follow up at all, and the small number who follow up so relentlessly it becomes a reason not to hire them. The optimum sits in the wide, underpopulated middle — and the interesting question is why, when the middle is this wide and this cheap to occupy, almost everyone defaults to silence. ## Why the follow-up works at all A follow-up doesn't make a weak application strong. What it does is more mundane and more reliable: it moves you back to the top of a pile, it signals genuine interest in this role rather than any role, and it gives a busy recruiter a low-effort way to respond — replying to a message is easier than initiating one, and hiring processes stall far more often from inertia than from decisions. There is also a selection effect. So few candidates follow up competently that doing it at all is differentiating. It's one of the rare places in a job search where the bar is on the floor. ## The timing that respects the process **After applying:** wait about a week, then send one short note. Earlier reads as impatience — the first days are usually queue time, and a nudge can't speed up a queue. If the posting names a closing date, wait until after it. **After an interview:** a brief thank-you within a day. Not flattery — a one-line restatement of why the conversation confirmed your interest, plus anything you promised to send. Its real function is keeping the thread warm while decisions happen. **After "we'll know by Friday":** let Friday pass, then one polite check-in. Deadlines slip constantly and harmlessly inside companies; your check-in is how you find out the slip happened, instead of reading silence as rejection. **After that:** one more, spaced by a week or two. Two unanswered follow-ups on the same thread is the practical ceiling. Past it, silence is the answer, and the correct response is to spend the energy on live threads. ## What the message should say Short enough to read in full inside a notification banner. Three sentences covers it: which role, one concrete reason you fit or one thing from the conversation that stuck, and a clear, answerable question — "is there anything you need from me?" beats "just checking in", because it invites a reply rather than an acknowledgement. Things that reliably hurt: guilt ("I was hoping to hear back by now"), pressure ("I have other offers", when deployed as a bluff), bulk-blast phrasing that obviously went to twenty recruiters, and length. Every additional paragraph reduces the chance of the message being read to its end, and the ask is at the end. ## Why nobody actually does this Not because it's hard — because of what it costs beyond the writing. Each follow-up requires remembering that a thread has gone quiet, deciding that enough time has passed, finding the correspondence, reconstructing the context (which CV did they get? what did we discuss?), and then composing something that doesn't sound needy — which, mid-search, with rejection ambient, is emotionally more expensive than it looks. Multiply by fifteen live applications, each on its own clock, and follow-up becomes a scheduling problem wearing a writing problem's clothes. Memory fails first; the follow-ups silently stop; and the applications most likely to be recoverable — the stalled ones — are exactly the ones that go unchased. This is the profile of a task worth automating: formulaic, low-stakes per instance, high-value in aggregate, and bottlenecked on memory and activation energy rather than skill. ## What automation should and shouldn't touch The division that works: the machine keeps the clocks and writes the drafts; you approve what leaves. Quarrion ties every recruiter thread to its application record, notices when a thread has sat quiet past the sensible interval, and drafts the follow-up with the context already in it — the role, the history, the artefacts that were sent. What it doesn't do is send anything by itself. Approving a good draft takes seconds, which is cheap enough that the follow-ups actually happen — but the words still pass through your judgement, which is what keeps them yours. The alternative — full automation, messages firing on schedules with nobody looking — belongs to the same family as application-blasting: it works until a recruiter notices, and recruiters notice. A follow-up is valuable precisely because it signals a person is attending to this thread. Remove the person and you're left with a cron job doing an impression of interest, which is worse than silence. Follow-up is the rare part of a job search where the leverage is high, the skill required is modest, and the only real obstacle is logistics. Fix the logistics and the middle ground between silence and pestering stops being underpopulated — at least by you. --- # Ghost jobs: why so many listings are dead, and how to stop applying into them Canonical URL: https://quarrion.ai/blog/ghost-jobs-and-dead-listings > A large share of job adverts are filled, paused or were never real. Where ghost listings come from, the signals that give them away, and how liveness checking works. Published 2026-08-18 · 4 min read Some of the applications you sent last month went to jobs that did not exist. Not scams — just adverts for roles that were already filled, quietly paused, or never had headcount behind them in the first place. Job seekers call them ghost jobs, and they are the single most demoralising feature of the modern job search, because they convert real effort into guaranteed silence and leave you wondering what you did wrong. Usually, nothing. There was no one reading. ## Where ghost listings come from Most ghost jobs are not malicious. They are exhaust from how hiring pipelines actually run. **The role was filled and nobody closed the advert.** Closing a posting on five boards is a manual chore that lands on someone with no incentive to do it promptly. The advert lingers for weeks after the offer was signed. **The posting is syndicated and copies outlive the original.** Company career pages feed aggregators, which feed other aggregators. The company closes the source; the copies persist downstream, sometimes for months, each one still accepting applications into the void. **Evergreen requisitions.** Large employers keep permanent adverts open for roles they hire continuously — or want a pipeline for. Your application goes into a pool, not a process. There may be no live vacancy behind it this quarter. **Hiring froze after the advert went up.** Budgets move faster than job boards. The recruiter may not even be allowed to say the role is paused. **Market-signalling and CV-harvesting.** The genuinely cynical minority: adverts posted to look like growth, benchmark salaries, or collect CVs for later. Rarer than the folklore suggests, but real. ## The signals worth checking by hand If you're vetting listings yourself, a few checks catch a lot: - **Age.** A posting up for a long time is either evergreen or dead. Boards that hide posting dates are hiding the single most useful field they have — which is worth remembering when choosing where to search. - **Cross-check the source.** Find the same role on the company's own career site. If it's on an aggregator but gone from the source, it's a stale copy. - **Reposting cycles.** The same role appearing, vanishing and reappearing for months without anyone apparently being hired is a pipeline advert. - **Vagueness.** Real vacancies name a team, a manager's remit, or a concrete problem. Adverts that read like a generic description of the profession are often not attached to a specific seat. The trouble with all four checks is arithmetic. Five minutes of vetting per listing across dozens of listings a week is hours of work — spent on the postings you *didn't* apply to. It's real diligence with an invisible payoff, which is exactly the kind of work people rationally skip. ## What automated liveness checking does This is a job for software, because the strongest signal is mechanical: is the posting still open at its source, and does it still accept an application? A listing that has been taken down, a form that has been disabled, an applicant tracking system that answers "this role is no longer accepting applications" — these are checkable facts, not judgement calls. Quarrion probes every listing for liveness before it ever reaches you, precisely because this check is cheap for a machine, expensive for a person, and binary in a way most job-search questions are not. A dead listing filtered out at the door costs you nothing; the same listing discovered after you tailored a CV for it costs you an evening and a small piece of morale. Filtering dead listings also quietly improves everything downstream. Match scores are only worth computing on roles that exist. Application counts only mean something if the applications had somewhere to land. A pipeline board full of applications-to-nowhere is not a record of a job search; it's a record of a leak. ## The part you still own No liveness check can see inside a company. A posting can be technically open and practically decided — an internal candidate, a favoured referral, a req the team expects to lose at the next budget review. Automation removes the mechanically dead listings, which is a large share of the problem; it cannot promise that every live listing is winnable. What you get back is proportion. When the dead adverts are gone, silence starts to mean something again — it's feedback about your application, not noise from a void. And the hours you were spending on diligence for roles that no longer existed go back where they belong: on the applications that have a human being at the other end. --- # How AI job matching actually works — and when a match score means anything Canonical URL: https://quarrion.ai/blog/how-ai-job-matching-actually-works > Keyword overlap, embeddings and LLM judgement produce very different match scores. What each method can and cannot see, and how to tell which one you're getting. Published 2026-08-18 · 4 min read Every job tool now shows you a match score. The number looks the same everywhere — a percentage, a grade, a coloured ring — but the machinery behind it varies enormously, and the machinery is what decides whether the number is worth anything. There are three broad ways to compute "how well does this person fit this role", and they fail in different ways. ## Method one: keyword overlap The oldest and still the most common. Extract terms from the posting, extract terms from the CV, count the intersection. The posting says Python, your CV says Python — a point. Say it more times, more points. This is cheap, fast, and explains most of the match scores you have seen. It is also vocabulary matching, not fit matching. It cannot distinguish a staff engineer from a graduate when both CVs contain the same technology names. It scores adjacent experience — the thing you did that isn't named in the posting but is obviously the same skill — as zero. And it is trivially gamed, which is why the folk advice about stuffing your CV with the posting's phrasing exists. If a tool's score moves dramatically when you paste in synonyms, this is what you're using. ## Method two: embeddings The step up. Both the posting and your CV are converted into vectors — long lists of numbers encoding meaning — and similarity is measured geometrically. Embeddings do capture the fact that "built ETL pipelines" and "data engineering" are the same idea in different words, which fixes keyword matching's blindness to phrasing. What embeddings still can't do is reason. Similarity is symmetric and requirements are not: a posting that requires five years of production experience is semantically close to a CV describing one internship in the same stack, and the vectors will happily report the resemblance. Embeddings answer "are these two documents about the same topic?" — which is necessary, and genuinely useful for surfacing candidates from a large pool — but the question you care about is "should this person spend an evening applying?", and that's a different question. ## Method three: an actual reading The expensive option: have a language model read the posting and the CV and make the comparisons a recruiter would. Which of the fifteen listed technologies are genuinely required and which are wishlist. Whether the seniority matches. Whether a gap is bridgeable ("has done the same job in a different stack") or fundamental ("has never done this job"). Whether the role fits where the candidate is trying to go, not just where they've been. This is the only method that produces a score you can argue with, because it can produce reasons alongside the number. It is also too expensive to run on every posting on the internet, which points at the architecture that actually works in practice. ## Why the practical answer is a pipeline Quarrion runs the three methods as tiers rather than picking one. Cheap rule-based filtering discards the obviously wrong first — wrong country, wrong seniority band, a duplicate of a listing already seen. A fast AI pass scores what survives. Only the roles that clear that bar get the expensive treatment: a deeper research pass that reads properly, checks the details and tags what it found. The economics are the point. Spending real reasoning on four hundred postings is wasteful; spending it on the thirty that survived triage is how you afford to do it well. The tier structure is what lets the final scores come from actual reading rather than word-counting, without the cost making the product impossible. ## How to evaluate any match score Whatever tool you're using, three questions separate a meaningful score from a decorative one: **Does it show its reasoning?** A bare number is unfalsifiable. If the tool can say *why* — required skill covered, seniority aligned, salary band plausible — you can spot when it's wrong and correct it. If it can't, you'll trust it until the first obviously bad match, then never again. **Does it know the difference between required and nice-to-have?** Show it a posting listing fifteen technologies. If a CV covering the four load-bearing ones scores worse than a CV name-checking ten peripheral ones, it's counting words. **Does it penalise seniority mismatch?** Search a senior title with a junior profile. A pure similarity engine will cheerfully rank the senior roles at the top, because the vocabulary matches perfectly. A tool that reads will not. A match score is a claim about your future time: apply here, not there. The machinery behind the claim decides whether following it is delegation or gambling — so it's worth the five minutes it takes to find out which machinery you're trusting. --- # How many jobs should you apply to? Fewer than the internet says Canonical URL: https://quarrion.ai/blog/how-many-jobs-should-you-apply-to > The applications-per-day question has the wrong unit. Why volume targets backfire, what actually limits a job search, and where the freed-up hours should go. Published 2026-08-18 · 4 min read Search this question and you'll find confident numbers: ten a day, fifty a week, a hundred a month. The numbers disagree with each other, which is the first clue that they're made up. The second clue is the unit. Applications per day treats a job search like piecework, where output scales with repetitions — and everything downstream of the submit button says it doesn't. ## What a volume target optimises Set a daily quota and you will hit it, because the quota is the only thing being measured. What bends to make that possible is everything the quota doesn't see. Reading the posting properly goes first. Tailoring goes second — a quota rewards the untailored application, which takes a tenth of the time and produces a tenth of the signal. Judgement about fit goes last: by application eight of ten, the marginal posting isn't one you chose, it's one that was still open in the tab. The applications this process emits are the ones recruiters describe reading constantly: technically responsive to the posting, obviously written for no posting in particular. They don't fail because a filter catches them. They fail because a person reads two lines and moves on — and each one taught you nothing, because nothing about "generic application, no reply" distinguishes the role, the CV, or the market as the cause. There's a second-order cost, too. Every application is a thread that can come back: a screening call, an assessment, a recruiter with questions about a role you can't remember applying to. Volume without judgement doesn't just lower reply quality — it books your own calendar with the replies you least wanted. ## The real constraint is attention, not throughput A job search has one genuinely scarce input: focused attention. The evening you have after work holds maybe two properly done applications — read the posting twice, decide what to emphasise, tailor, write something specific, check it. That's not an argument for a quota of two. It's the observation that attention spent on a badly matched role is attention removed from a well matched one, which reframes the entire question. The number that matters isn't how many you apply to. It's how good the *worst* role you applied to was. That number is set upstream of the application, at the filtering step — the unglamorous work of reading many postings and discarding most. Done honestly it consumes more time than applying does, which is exactly why quota-driven searches skip it. Skipping it doesn't remove the filtering; it moves the filtering to the employer's side, where it runs at two lines per rejection and returns no information to you. ## Where automation actually belongs in this The tempting automation is the submit button — it's the most repetitive step, so it looks most automatable. But automating submission scales the part that was never the constraint, and floods the market with exactly the applications that don't work. The automation that changes the economics sits earlier and later. Earlier: discovery and filtering — scanning the boards, discarding the dead and the mismatched, scoring what's left against your actual history so the evening starts at the shortlist instead of the haystack. This is where Quarrion puts the machine reading: it reads the four hundred so your attention starts at the twenty. Later: the clerical wrapper around each application you *chose* — the form-filling, the tracking, the follow-up drafts — with an approval gate, so nothing leaves without a person having decided it should. The judgement stays where the accountability is. Notice what this does to the original question. With filtering handled, the sensible application count stops being a discipline target and becomes an output: it's however many roles genuinely cleared the bar this week. Some weeks that's three. In a hiring surge in your specialism it might be fifteen. Both are correct answers, because the bar, not the count, is what's being held constant. ## A better weekly shape If a number helps as scaffolding, aim it at inputs you control: every open role in your target space *seen* and triaged this week — by software if possible, by a fixed skim hour if not; every role that cleared your bar applied to properly, with materials arranged for that reader; every live thread followed up on schedule. Then let the application count fall where it falls, and judge the week by pipeline movement — screens booked, interviews progressing — rather than by submissions. "How many should I apply to?" is usually anxiety wearing arithmetic's clothes — the search feels stalled, and a bigger number feels like doing something. The honest answer is that the number was never the lever. The bar is the lever. Set it with your own judgement, hold it when the week feels slow, and spend the anxiety on the follow-ups instead; they're cheaper than applications and recover stalled threads more often. --- # How to track job applications without the spreadsheet dying by week three Canonical URL: https://quarrion.ai/blog/how-to-track-job-applications > Why application-tracking spreadsheets always decay, what a working system has to record, and the point at which tracking is worth automating alongside the rest. Published 2026-08-18 · 4 min read Everyone's job-search tracking system starts the same way: a spreadsheet, created with great optimism, with columns for company, role, date and status. And almost everyone's spreadsheet dies the same death, somewhere around week three, when the search gets busy enough that updating the tracker starts competing with the activities it was supposed to organise. The record decays exactly when it would start being useful. This isn't a discipline failure. It's a design failure, and it's worth being precise about why — because the fix follows from the diagnosis. ## Why the spreadsheet dies **Every update is manual, and updates are where the value lives.** A tracker is only as good as its worst week. The week you have two interviews, a take-home and a recruiter chasing you is the week you stop logging — and it's also the week the record mattered most. **Status isn't one column.** "Applied" hides a dozen real states: applied and acknowledged, applied into silence, screening booked, take-home pending, awaiting feedback, ghosted-but-maybe-not. A single status cell forces you to flatten states that need different actions, so the tracker stops telling you what to do next — at which point it's an archive, not a tool. **The correspondence lives somewhere else.** The actual substance of your search — who said what, when, and what you promised — is in your inbox and your LinkedIn messages. A spreadsheet holds a pointer to a conversation, not the conversation. When a recruiter calls about "the role you applied for", the spreadsheet tells you a company name and a date; it can't reconstruct what was sent, which version of your CV they have, or what you last told them. **Nothing chases you.** The highest-value act in application tracking is the follow-up — the polite nudge at the right moment, the check-in after an interview. A spreadsheet knows the date of your last touch and will never once remind you. ## What a working system has to record Whatever tool you use — and a spreadsheet *can* work for a small, slow search — the record needs to hold, per application: - **The exact artefacts sent**: which CV version, which cover letter. If you tailor per role (you should), "my CV" is not one document, and interviews go noticeably worse when you can't remember which framing they read. - **The full correspondence thread**, attached to the application rather than scattered across inboxes. - **A real status**, granular enough to imply the next action. - **The next action and its date** — yours ("follow up Friday") or theirs ("they said end of week"). - **The posting itself**, saved at apply time. Listings get taken down, and walking into an interview without the job description is avoidable. Notice how much of this is clerical capture of things that already happened in other systems. That's the tell that it's automatable. ## The pipeline-board shape The kanban framing — columns for applied, screening, interviewing, offer — beats the spreadsheet for one structural reason: it makes state transitions visible instead of cell edits. You see where every application sits, where things are piling up, and where the pipeline is leaking. Applied-and-silent for two weeks is visually distinct from actively-interviewing, and different columns imply different actions. Quarrion uses this shape, with the mechanical parts wired in: applications land on the board automatically when they're submitted (the agent submitted them, so there's nothing to transcribe), the recruiter correspondence sits on the application record in a unified inbox rather than in a parallel universe, and follow-ups are drafted and tied to the thread they belong to. The tracking is a by-product of the pipeline running, which is the only reliable way tracking ever happens — a record that maintains itself doesn't decay in week three, because nobody is maintaining it. ## When tracking starts paying for itself Whatever the system, the payoff compounds with volume. At five live applications, memory works. At fifteen, cross-talk starts: you re-apply to a company that rejected you last month, or blank on a recruiter's question about a role you can't place. Past twenty, an untracked search isn't merely disorganised — it produces actively worse outcomes, because follow-ups don't happen, and follow-ups are among the cheapest actions in the entire search. There's also a quieter benefit: an accurate pipeline is the only honest measure of how the search is going. Applications-sent is a vanity number. What a real record shows is where things stall — plenty of applications but no screens says one thing; screens that never convert to interviews says another. You can't fix a leak you can't see, and seeing it is the whole point of writing things down. --- # What is Quarrion? An honest tour of the pipeline Canonical URL: https://quarrion.ai/blog/what-is-quarrion > What the agent actually does at each stage — discovery, scoring, research, tailoring, applying and follow-up — and the parts it deliberately leaves to you. Published 2026-08-18 · 4 min read Quarrion is an AI agent that runs a job search. The one-line version: it scans, scores, tailors and applies — you interview. That line is doing a lot of compression, so this post walks through what actually happens at each stage, and, just as importantly, what the agent deliberately does not do. ## Discovery: reading the boards so you don't Openings are scattered across more than a hundred places — the big aggregators, the specialist boards, and the applicant tracking systems that host company career pages directly: Ashby, Workable, Greenhouse, Lever and their peers. The good roles are often only in one of them, and frequently it's the company's own page, which nobody checks daily. Quarrion checks them daily. This is the least glamorous part of the product and arguably the most valuable per hour saved, because checking a hundred sources is exactly the kind of high-volume, zero-judgement work software should absorb. Before any listing reaches you, it is probed to confirm the role is still open. Ghost listings — roles that were filled, cancelled or never real — are the most common complaint about job boards, and applying into one costs you an application's worth of effort for a guaranteed silence. Dead listings get filtered out at the door. ## Scoring: judgement, not keyword counting Every discovered role is scored against your CV. The scoring runs in tiers: cheap rule-based filtering first (wrong country, wrong seniority, duplicate of something already seen), then an AI pass that reads the posting the way a person would — what the role actually needs, whether your experience covers it, whether the gap is bridgeable or fundamental. Roles that score well enough get a deeper research pass: the agent digs into the role and tags what it found, so a strong match arrives with context rather than just a number. The score always shows its working. That is a design decision, not a flourish: a number you can't interrogate is a number you can't disagree with, and you will sometimes disagree with it. The breakdown is what lets you correct the agent rather than abandon it. ## Tailoring: grounded in what you've actually done For roles you want to pursue, Quarrion tailors your CV and drafts a cover letter per role. The tailoring is grounded in a structured profile built from your own CV — a knowledge graph of your actual roles, skills and accomplishments — which means the rewriting works by selection and emphasis, not invention. It can foreground the project that matters for this posting. It cannot give you three years of Kubernetes you don't have, because nothing in the source material says you do. Made-up skills are a documented failure mode of AI CV tools, and the fix is architectural: if the generator can only draw from your verified history, the hallucination has nowhere to come from. ## Applying: the agent fills, you approve The agent fills real application forms — the tedious, repetitive part — and then queues the application for your approval. Nothing is submitted until you say so. This gate is the load-bearing difference between Quarrion and the spray-and-pray blasters. An application an employer reads as personal should have had a person behind it at least once, at the moment of decision. Automating everything up to that decision saves the hours; keeping the decision keeps you honest, and keeps your name off the pile of obviously-automated applications that recruiters learn to skip. ## After the application: the loop most tools drop Applying is the middle of a job search, not the end. Quarrion tracks every application on a pipeline board from applied to offer, drafts recruiter replies and follow-ups in a unified inbox — each tied to the application it belongs to, so you never answer a call about a role you can't place — and helps with interview preparation and offer negotiation when things progress. There is also a salary layer: city-level benchmarks adjusted for cost of living and inflation, with live exchange rates, so "is this offer good?" gets a real answer rather than a guess. ## What it deliberately doesn't do No application is sent without your approval. Your CV is not rewritten with skills you don't have. The agent doesn't apply to hundreds of roles to manufacture a feeling of progress — the scoring exists precisely so most roles never take up your attention at all. The premise of the whole product is that the bottleneck in a job search was never how fast you could submit. It was reading four hundred postings to find the twenty worth an evening. Quarrion reads the four hundred. You read the twenty, approve what goes out, and take the interviews. There is a free tier with no card required, which gets you scored roles in the first session. The paid tiers add volume, auto-apply and tailoring. Either way, the shape of the deal is the same: the agent does the reading, the drafting and the form-filling; the decisions stay yours. --- # When an agent runs your job search, what's left for you to do? Canonical URL: https://quarrion.ai/blog/what-to-do-while-the-agent-searches > Delegating discovery, filtering and drafting frees most of a job search's hours. The highest-return places to reinvest them: skill gaps, interviews and your network. Published 2026-08-18 · 4 min read The traditional job search spends its hours on logistics: scanning boards, reading postings, filling forms, chasing threads. When an agent absorbs that layer — discovery, filtering, tailoring, tracking, follow-up drafts — a strange thing happens: the search suddenly has spare capacity, and no habit for spending it. The hours used to be allocated by necessity. Now they have to be allocated by choice, and the choice is where searches quietly succeed or stall. Here's the honest map of where reinvested hours pay, roughly in order of return. ## Close the gap the market keeps naming A filtered, scored job search generates a by-product that unfiltered searching never does: evidence. When every posting in your space has been read and scored against your actual history, patterns surface — the same requirement keeps appearing in roles you want, and it keeps being the thing you're missing. Quarrion surfaces this directly as skill-coverage insights across your matched roles: what the market you're targeting keeps asking for that you don't yet have. This turns the vague guilt of "I should upskill" into a shopping list with prices. One missing skill that appears in most of your target roles is worth a focused fortnight; a skill that appears rarely can be ignored without guilt. The job search becomes, in part, a curriculum — and study aimed by market evidence beats study aimed by trend articles every time. ## Rehearse interviews like they're the actual exam Because they are. Once discovery and applications are handled, the conversion points that remain are conversations — screens, technical rounds, final panels — and they're the least automatable part of the entire pipeline, because they're the part where a company is buying *you*. Interview performance responds to rehearsal faster than almost any skill in the search. Saying your answers out loud, discovering which stories ramble, tightening the explanation of why you left, practising the walk-through of your best project — an hour of this moves outcomes more than an evening of extra applications. Preparation compounds, too: the story bank you build for one interview is most of the preparation for the next. This is also where falling application logistics should show up as *rising* interview quality — if the agent saves you six hours and none of them reach interview prep, the reallocation failed. ## Work the channel that bypasses the pile Referred candidates skip the coldest part of the funnel — that's not news. What changes when logistics are delegated is that you finally have the hours to act on it. Networking done honestly isn't schmoozing; it's maintenance of real relationships: the former colleague at the company you just applied to, the meetup in your specialism, the considered comment in the community where your field actually talks. A warm introduction doesn't replace the application — it changes which pile the application lands in. The sequencing matters: this channel works best *alongside* an active pipeline, because "I've just applied for the platform role — would you be willing to refer me?" is an easy, concrete ask, while "let me know if you hear of anything" is homework you've assigned to a friend. ## Do the diligence that changes decisions Late-stage time is decision time: is this company somewhere you want three years of your life to go? Glassdoor patterns (themes, not individual reviews), the financial signals — funding, layoffs, growth — the trajectory of people who held this role before you, what former employees say when asked directly. This research is unautomatable in the part that matters, because the question isn't "what are the facts?" but "do these facts fit *me*?". An afternoon here, before an offer call, is worth more than the same afternoon anywhere else in the pipeline — it's the difference between negotiating a job you understand and accepting one you don't. ## Rest, on purpose The unfashionable one. Job searching under logistics pressure produces a particular grind — always behind, always one more posting to check — and that grind leaks into interviews as flatness and into decisions as desperation. Part of the point of delegating the clerical layer is that evenings can actually end. A candidate who slept, exercised and did something unrelated to the search interviews noticeably better than one who spent the same hours refreshing job boards. Recovery isn't stolen from the search; it's part of the machinery. ## The reallocation is the strategy None of this is what people picture when they imagine automating a job search — the picture is usually "the machine does everything and I wait". The truer picture is a trade: the agent takes the hours that were logistics, and the search's outcome now depends on what you do with them. Skills, rehearsal, relationships, diligence, recovery — the human work was always the high-leverage part. It just used to lose its hours to the form-filling. --- # Do AI job application agents actually work? Canonical URL: https://quarrion.ai/blog/do-ai-job-application-agents-work > The honest answer depends entirely on which half of the problem the tool automates. Most automate the wrong half, and the difference shows up in your reply rate. Published 2026-08-10 · 5 min read "AI job application agent" now covers two completely different kinds of software, and the distinction matters more than any feature list. The first kind automates **submission**. You give it a search — a job title, a location, maybe a salary floor — and it fires applications at everything that matches, filling forms with a stored CV. The pitch is volume: 500 applications while you sleep. The second kind automates **judgement**. It reads the posting, compares it against what you have actually done, and decides whether the role is worth applying to at all. Submission is the last step, not the product. Both get called agents. Only the second one addresses why job searching is hard. ## Volume was never the bottleneck The case for spray-and-pray sounds reasonable. Applications are a numbers game, each one is mostly clerical, so automate the clerical part and multiply the numbers. It fails on both ends. At the employer's end, an application that obviously wasn't written for the role is easy to spot and cheap to discard. Recruiters read hundreds of these. The generic ones do not fail because a filter catches them — they fail because a person reads two lines and moves on. At your end, volume creates work you then have to do. Every application is a thread that might come back: a screening call, a take-home, a recruiter asking about a role you don't remember applying to. Five hundred applications is not five hundred lottery tickets. It's an inbox you now have to manage, mostly containing roles you were never a good fit for. The bottleneck was never how fast you could submit. It was working out which twenty of the four hundred open roles are actually worth your afternoon. ## What "scoring" has to mean to be worth anything Every tool in this category claims to match you to jobs. The claim is load-bearing, so it's worth being precise about what would make it true. Keyword overlap is the cheap version, and it's what most matching amounts to. The posting says Python, your CV says Python, that's a point. This ranks by vocabulary. It cannot tell the difference between someone who used a tool once on a university project and someone who ran it in production for three years, because both CVs contain the same word. Useful scoring has to reckon with things keyword overlap cannot see: - **Seniority.** A staff-level posting and a graduate posting share almost all their vocabulary. They are not the same job and you are not right for both. - **The difference between required and nice-to-have.** Most postings list fifteen technologies and genuinely need four. Which four is a judgement about the role, not a count of words. - **Adjacent experience.** The thing you did that isn't named in the posting but is obviously the same skill. Keyword matching scores this as zero. - **Direction.** A role can be a good match for your CV and a bad match for where you're trying to go. Those are different questions and a tool that only answers the first will happily march you sideways. This is why the honest version of the pitch is not "we apply to more jobs." It's "we read more jobs than you have time to, so you only read the ones worth reading." ## Where automation genuinely helps Having argued against automating submission, here is the case for automating several things around it. **Discovery.** Openings are scattered across job boards, company career pages and applicant tracking systems, and the good ones are often only on one of them. Checking a hundred sources daily is exactly the kind of tedious, high-volume, zero-judgement work software should do. Quarrion scans 100+ boards for this reason. **First-pass filtering.** Reading four hundred postings to find twenty is not a good use of a human evening. Reading the twenty is. **Tailoring.** Rewriting a CV per application is the step people skip, because it's genuinely tedious and the payoff is invisible until it isn't. It's also the step most worth doing, which makes it the best possible thing to automate. **Correspondence.** Recruiter replies, follow-ups, scheduling. Necessary, low stakes, formulaic. Draft them automatically and approve them in seconds. Notice the pattern: automate the work that is high-volume and low-judgement, and give the human the decisions. Submission automation inverts this — it hands the judgement to the machine and keeps the tedium for you, because you still have to deal with everything that comes back. ## Questions worth asking any tool in this category - **Does it show you why it scored a role the way it did?** A number with no reasoning is not something you can disagree with, and you will need to disagree with it. - **Does anything leave without your approval?** There is a real difference between a tool that drafts and one that sends. Know which you have bought. - **What happens to your CV?** It is the most sensitive document most people own. Where it's stored, which third parties process it, and whether you can delete it should all be answerable without emailing anyone. - **Can you see what it applied to?** If you cannot reconstruct what was sent on your behalf, you cannot answer a recruiter who calls about it. ## The honest summary Do they work? A tool that automates submission mostly converts your time into someone else's spam folder. A tool that automates discovery, filtering and tailoring gives you back the evening you were spending on job boards and puts better applications in front of better-matched roles. The word "agent" doesn't tell you which one you're looking at. What it automates does. --- # Tailoring your CV per application: what it actually means Canonical URL: https://quarrion.ai/blog/tailoring-your-cv-per-application > Not rewriting it from scratch, and not stuffing it with keywords. A practical account of which parts of a CV should change per role, and which never should. Published 2026-08-08 · 4 min read Everyone is told to tailor their CV for each application. Almost nobody is told what to change, so it usually collapses into one of two failure modes: pasting the job title at the top and calling it tailored, or rewriting the whole document each time until the effort makes you stop applying. The useful version is narrower than the second and much more than the first. ## What tailoring is not **It is not keyword stuffing.** The folk wisdom says applicant tracking systems reject CVs that don't contain enough matching terms, so you should smuggle in every phrase from the posting. This gets the mechanism wrong. Most ATS software does not auto-reject on keyword density — it stores, sorts and searches applications. A recruiter runs a search, skims results, and reads a shortlist. Keywords matter because they affect whether you surface in that search, not because a robot is scoring you out of a hundred. And a CV written for a keyword parser reads badly to the person who opens it, which is where the actual decision happens. **It is not a new CV each time.** If tailoring takes an hour, you will do it for three applications and then stop. A process you abandon is worse than a lighter one you sustain. ## The parts that should change Think of a CV as having a fixed spine and a variable surface. The spine is factual: where you worked, when, job titles, education, the technologies you genuinely used. This should never change, for the obvious reason that it's a record of things that happened, and for the practical one that inconsistencies across versions surface in background checks and second interviews. The surface is editorial: what gets emphasised, what order things appear in, which of your accomplishments make the cut, and how each is phrased. This is what should change, and it's a smaller job than rewriting. Concretely, four things: **1. The opening summary.** The two or three lines at the top are the highest- leverage text in the document, because they're the only part guaranteed to be read. They should describe the person the posting is looking for, if that person is honestly you. **2. Which bullets survive.** Most roles you've held produced more accomplishments than fit on a page. Which ones you keep should depend on the role you're applying for. A backend-heavy posting and a platform-heavy posting can pull entirely different bullets from the same job. **3. The vocabulary for the same fact.** If you built a system that moved data between services on a schedule, that is "ETL pipelines" on a data posting and "service integration" on a platform posting. Same work, and the posting tells you which framing the reader already has in their head. This is where keyword alignment is legitimate — you're describing real work in the reader's language, not inventing it. **4. Ordering.** Putting the most relevant role first within a section, or the most relevant bullet first within a role, costs nothing and changes what someone sees in the six seconds they spend on the first pass. ## The parts that must not change Worth stating explicitly, because tailoring shades into fabrication more easily than people expect: - Dates, titles and employers. - Technologies you have not used. Reframing "wrote SQL queries" as "data engineering" is a stretch; adding Kafka because the posting mentions it is a lie you will be asked about in a technical interview. - Quantified claims. If you don't know the number, don't invent one. "Reduced latency" is weaker than "reduced p99 latency by 40%" and infinitely better than a fabricated 40%. - Anything you would not be comfortable being asked to expand on for ten minutes. That is the actual test, and it's the one interviews apply. ## Making it sustainable The reason tailoring gets skipped is that it competes with applying to more roles, and volume feels like progress. Two things make it survivable. **Keep a master document.** Not a CV — a longer file with every role, every project and every accomplishment, including the ones that never fit on a page. Tailoring then becomes selection from a menu rather than composition from a blank page. This is most of the work, and you do it once. **Be honest about which applications deserve it.** If a role is a marginal fit, a tailored CV will not rescue it, and the time is better spent on one you'd actually take. Tailoring is a multiplier on a decent match, not a substitute for one. This is where automation earns its place. Selecting the relevant bullets and rephrasing them for a specific posting is a bounded, repetitive editorial task with a clear input and a clear output — the kind of thing software does well. Deciding which roles are worth applying to, and checking that the result is true, remain yours. Quarrion drafts the tailored version per role and gives it to you to approve, which is the right split: the machine does the tedious part, and a human signs off on the claims. ## A test for whether you've done it Read the opening summary. If you could paste it onto an application for a different role at a different company without changing a word, you haven't tailored the CV — you've reformatted it. --- # Automating a UK job search without annoying employers Canonical URL: https://quarrion.ai/blog/automating-a-uk-job-search > Which parts of a UK job hunt are safe to automate, which are not, and the specifics that differ here — ATS platforms, right-to-work questions and salary bands. Published 2026-08-06 · 4 min read Most advice about automating a job search is written for the US market, and enough of it transfers badly that it's worth going through what actually differs in the UK — and where the line sits between automation that helps and automation that burns a company you might want to work for later. ## The line A useful rule: **automate everything up to the point where a human at the other end would feel misled.** Scanning boards, filtering, drafting, tracking, reminding — all invisible to the employer and none of it changes what they receive. Automating those is uncontroversial. Sending an application the employer will read as personal, when nobody looked at the posting, crosses the line. Not because it's against the rules, but because it works exactly once per company. Recruiters remember the source of a bad application, and the UK market in most specialisms is smaller than people assume. The same twenty companies and the same handful of agencies will keep coming up. The practical version: automate up to a draft, decide yourself whether to send. ## What's different about applying in the UK **The ATS mix.** Greenhouse, Lever and Workday dominate at larger employers, and below that you hit a long tail of smaller systems and plain email applications. Any tool claiming to auto-apply everywhere is claiming to handle a very long tail — worth checking what it actually does when it hits a form it doesn't recognise. Failing loudly and handing it back to you is the correct behaviour. Silently marking it applied is not. **Right-to-work questions.** Nearly every UK application asks some version of "do you have the right to work in the UK?" and "will you now or in the future require sponsorship?" These are legally significant, and the second one is not the same question as the first. Getting them wrong — in either direction — wastes everyone's time or misrepresents you. This is the single field most worth checking on any form that was filled for you. **Salary bands.** UK postings disclose salary far less consistently than US ones, and there is no equivalent of the state-level pay transparency laws that have made US ranges more common. A lot of listings say "competitive." Filtering strictly on an advertised salary floor will silently discard a large share of the market, including roles that would have met your number. **Agencies.** A significant share of UK listings are posted by recruitment agencies rather than employers, sometimes several agencies advertising the same role, sometimes without naming the company. This has two consequences for automation: you will see duplicates that aren't obviously duplicates, and an application may go to an intermediary who will contact you about entirely different roles. **Notice periods.** One and three month notice periods are common here, longer at senior levels. This changes the shape of the search — the pipeline is slower and starts earlier than the US equivalent — and it makes tracking what stage each application is at more valuable, because the timeline is long enough to lose track. ## What to automate **Discovery, without reservation.** Checking many sources daily is pure repetitive work with no judgement in it. This is the clearest win, and it's why Quarrion scans 100+ boards rather than asking you to. **Deduplication.** Given the agency situation, the same role reaching you three times under three names is normal. Collapsing those is unglamorous and saves real time. **First-pass filtering.** Something has to reduce a few hundred daily postings to a readable number. Just make sure you can see why something was filtered out — an opaque filter will quietly discard a role you'd have wanted, and you'll never know. **Drafting.** Tailored CVs, cover letters, recruiter replies, follow-ups. All formulaic, all tedious, all improved by having a draft to react to rather than a blank page. **Tracking.** With three-month notice periods and multi-stage processes, the state of twenty applications is genuinely hard to hold in your head. ## What not to automate **The final send, at least at first.** Approve each one until you trust the drafts. This is a temporary cost that buys you the ability to catch a misinterpreted posting before an employer sees it. **Right-to-work and salary expectation fields.** Check every time. These are the ones with consequences. **Anything addressed to a named person.** If the posting names the hiring manager, that is an invitation to write something specific, and it's the cheapest signal of genuine interest available to you. **Roles you'd genuinely take.** The best five applications in your search deserve an hour each. Automation should be buying you that hour, not spending it. ## A reasonable setup 1. Automated discovery across as many sources as you can stand. 2. Automated scoring against your actual CV, with visible reasoning. 3. You read the shortlist — the twenty, not the four hundred. 4. Automated drafting of the CV and cover letter for the ones you pick. 5. You read the draft, check the right-to-work answers, and send. 6. Automated tracking and follow-up reminders. Steps 1, 2, 4 and 6 are machine work. Steps 3 and 5 are the job. A tool that takes 3 and 5 away from you is not saving you time — it's making decisions you will have to live with, faster than you can check them.