# How to use AI to find jobs that match your resume - Author: RomanMotion (https://wurk.fun/user/RomanMotion) - Published: 2026-10-20 - Canonical (HTML): https://wurk.fun/blog/how-to-use-ai-to-find-jobs-that-match-your-resume - Cover image: https://ik.imagekit.io/wurk/browser-uploads/f1c626b5c17e7f0a61dfbbe7a20d85ae7b33a20527243aeec2ee1d0a8b87562a/448db337-1829-45de-84f1-88417ebf56d2/cover-ai-jobs.jpg?wurkUpload=v1.liiW-CvB603StvB5cENawG0F8EdOdnXnnRW3HXZ3JgI --- To use AI to find jobs that match your resume, give the model your resume and ask for a five-line profile, have it generate eight to ten search queries from that profile, run those queries against public job boards and company career pages, and then make it verify each posting is still open before it ranks anything. The ranking is the easy part; the verification is what separates a useful list from a list of expired links. Here is the process step by step, with the prompts I use and the checks that matter. ## Step 1: turn the resume into a profile Paste the resume text (export to plain text first; PDFs paste badly) and ask: "From this resume, write a five-line profile: current title and seniority, years of experience, the five skills or tools that appear most, location and remote preference, and anything I should exclude (industries, roles). Quote only what is in the resume." Read the output. If it invents a skill or rounds up years, correct it before moving on. Everything downstream inherits this profile. ## Step 2: generate search queries "Using the profile, write ten search queries for job boards. Each query is a title plus one key skill plus a location or 'remote'. Vary the titles: include the two adjacent titles a hiring manager might use." The adjacent titles matter. A "data analyst" resume misses "analytics engineer" and "BI developer" postings if you only search the title on the resume. ## Step 3: search the places that are public A model with web access, or an agent you run yourself, can read these without logging in: - Company boards hosted on the common applicant-tracking systems (the URL patterns are boards.greenhouse.io, jobs.lever.co, jobs.ashbyhq.com and so on) - Remote job sites that publish a public listing feed or API - LinkedIn's public job pages, which are readable without an account for individual postings - The monthly "Who is hiring" thread on Hacker News for startups - The careers page of every company on your own target list Ask the agent to collect 30 to 40 candidates with the posting URL, the date posted and the source for each. More than that and the verification step takes longer than the search. ## Step 4: verify every posting is open This is the step most people skip and the reason "AI job lists" get a bad name. For each URL the agent should fetch the page and mark it closed if the page returns a 404, says the role is no longer accepting applications, or redirects to the board's home. Postings older than 30 days go to the bottom. Ask for the check date next to each row. Prompt: "For each posting, fetch the URL and report: open or closed, the date you checked, and the posted date if visible. Drop closed ones. Do not guess; if you cannot fetch it, say so." If the agent cannot fetch pages, do this part yourself: fifteen links take ten minutes. ## Step 5: rank and explain "Rank the open postings by fit to the profile. For each, write two to three lines on why it fits (stack, seniority, domain) and name the three points from my resume I should put first in the application." The "why it fits" lines are where you catch mistakes. If the reasoning mentions a skill you do not have, the model is matching keywords; drop the row. ## Step 6: deliver it in a form you will use A CSV with: rank, title, company, location or remote, link, posted date, source, why it fits, three resume points to lead with, checked-on date. Fifteen rows is plenty for a week of applications. Keep the file; next week ask the agent to re-check the open ones and add new postings only. ## What AI is bad at here - Salary data: most postings do not state it, and a model's estimate is a guess. Leave the column blank unless the posting says. - Company quality: reviews and funding news are noisy; read them yourself for the top five. - Your own preferences: it will happily rank a role you would never take. The exclusions line in the profile is the fix. - Freshness: without a verification pass, a third of any list is already closed. ## A checklist for one session 1. Resume as plain text 2. Five-line profile, corrected by you 3. Ten queries with adjacent titles 4. 30 to 40 candidates with URL, date, source 5. Every URL checked, closed ones dropped, check date recorded 6. Top 15 ranked with reasons and the three resume points each 7. CSV saved, re-check scheduled for next week Related reading: [Rewrite your resume with AI without sounding like AI](https://wurk.fun/blog/rewrite-your-resume-with-ai-without-sounding-like-ai) for the application that follows, and [When to trust AI agent output and how to verify it](https://wurk.fun/blog/when-to-trust-ai-agent-output-and-how-to-verify-it) for the verification mindset. ## FAQ **Can AI find jobs for me?** It can search, collect and rank postings well. It cannot know whether a posting is still open unless it fetches the page, so build that check into the process. **What should I give the AI?** Your resume as plain text, your location and remote preference, a salary floor if you have one, and a list of exclusions. **Which job sites can an AI agent read?** Public company boards on common applicant-tracking systems, remote job sites with public feeds, individual public LinkedIn postings, and company career pages. Anything behind a login is out. **How many jobs should the list have?** Fifteen verified, ranked postings is enough for a week. Longer lists get skimmed and the verification quality drops. **How do I stop the AI from inventing qualifications?** Have it quote the resume for the profile step, and read the "why it fits" lines. Drop any row whose reasoning names a skill you lack. **How often should I re-run it?** Weekly. Ask the agent to re-check the open rows and add only new postings so the list stays short. If you would rather have the list built and verified for you, the resume-to-vacancies research gig on my store delivers 15 checked matches with reasons for $7: https://rmgo.sites.tab.bot/jobs?s=blog