Find Jobs That Match Your Resume with AI

Most job searches fail in the same place, and it is not the interview. It is the list. People spend weeks applying to roles that were never going to say yes, and conclude that the market is broken or that they are not good enough. Usually neither is true. The list was wrong.

This guide is the first half of a two-part method. Here we go from the resume you already have to a short, honest list of roles worth your time. In part two we go the other way: from a specific vacancy to a resume rewritten for it.

The order matters. Tailoring a resume to the wrong role is wasted effort no matter how well you do it. Find the right roles first.

What this costs and what you get

Time: about an evening to set up, then roughly an hour a week to run.

You will need: your current resume, any AI assistant, and one spreadsheet.

You end with: a written search brief, queries that return relevant roles on every board you use, a scoring rubric that sorts postings into apply, stretch and skip, and a pipeline that tells you whether your problem is targeting or the resume itself.

The seven steps

1. Extract your real signal from the resume
2. Write the search brief
3. Turn the brief into queries that boards understand
4. Score every posting against a rubric
5. Read what the posting is not saying
6. Verify before you spend anything
7. Keep a pipeline, and let it correct you
The prompts · Common questions

Why applying to more jobs makes things worse

The intuition is reasonable: more applications, more chances. In practice it breaks down for three reasons, and it helps to know which one is hurting you.

You cannot tailor at volume

•A tailored application takes 30–60 minutes of real thought
•At 200 applications that is a full working month you do not have
•So the applications get generic — and generic loses to specific every time

You learn nothing from silence

•A rejection tells you something; no response tells you nothing
•High-volume searches produce mostly silence
•Without signal you cannot correct course, so month two repeats month one

It costs you the roles you could win

•Attention is the scarce resource, not job postings
•Every hour on a role that was never a fit is an hour not spent on one that was
•The good applications end up rushed because the bad ones ate the time

The fix is not to apply less. It is to spend your applications where they can actually land. That requires knowing, before you apply, how close a given role is to what you can already evidence.

What AI is genuinely good at here — and what it is not

Be precise about the division of labour, because most disappointment with AI in a job search comes from asking it to do the part it cannot do.

Give this to the model

•Reading fifty job descriptions and extracting their real requirements
•Naming the roles your experience maps to, including titles you had not considered
•Turning a description of what you want into search queries that a job board understands
•Scoring postings against a rubric you wrote, consistently, without getting tired
•Spotting what a posting quietly says about the team — scope, seniority, stability

Keep this for yourself

•Deciding what you actually want, and what you will not accept
•Judging whether a company is somewhere you want to spend two years
•Anything a recruiter could check — dates, titles, numbers, employers
•The final call on whether to apply

The rule that keeps you out of trouble

AI reads and sorts. You decide and verify. The moment a model is deciding what you want, or writing claims you cannot back up in an interview, you have handed it the wrong job.

Step 1. Extract your real signal from the resume

What you do: paste your current resume into a model and ask it to describe you the way the market reads you — not the way you meant it.

Why this comes first: there is almost always a gap between the career you have had and the career your resume describes. You know you led a migration; the resume says "participated in infrastructure initiatives." You will search for the job you remember doing, while every automated filter reads the weaker version. Seeing the gap early is the whole point of this step.

What you ask for, specifically: the role families your experience maps to, the seniority band a stranger would place you in, the domains you can claim without stretching, the technologies that appear with real weight versus the ones mentioned once, and — most useful — the three things a reader would remain unsure about.

If you skip it: you will search using your own job title, which is often company-specific and matches nothing. People who were "Head of Platform" at a forty-person startup search for "Head of Platform" and find twelve roles, when the market has hundreds of Staff Engineer and Engineering Manager roles that fit them exactly.

What good output looks like

A list of five to eight concrete role titles, a seniority band with a one-line justification, two or three domains, and an honest note about what the resume fails to make clear. If the model returns flattery, ask it again for the version a sceptical recruiter would write.

Step 2. Write the search brief

What you do: write one page — yourself, not the model — that states what you are looking for and what you will refuse.

Why: every later step depends on this document. The queries come from it, the scoring rubric comes from it, and the decision to apply or skip comes from it. Without it, "is this a good role?" has no answer, and you will drift toward whatever is most recently posted.

What goes in it: the role titles from step one; the seniority band; company shape — stage, size, whether you want a team to build or a system to run; domain preferences and hard exclusions; location, timezone, and work-authorisation constraints stated plainly; a compensation floor, not a wish; and a short list of deal-breakers.

The part people get wrong: separating must-have from nice-to-have honestly. If everything is a must-have, nothing passes and you conclude the market is empty. Three to five must-haves is a workable number. Everything else goes in the second column and becomes a tiebreaker.

A test for the brief

Hand it to a friend along with three job postings. If they can pick the same one you would have picked, the brief is doing its job. If they cannot, it is too vague to filter anything.

Step 3. Turn the brief into queries that boards understand

What you do: give the model your brief and ask for search strings for each place you will actually look — a general engine, the job boards you use, and any company career pages you care about.

Why AI helps here: the same role is advertised under six different titles, and the one you know is rarely the most common. A model that has read a great many postings is good at producing the synonym set — the titles, the tools listed as proxies for the work, the phrases that indicate the seniority you want.

What to ask for: several query variants per board rather than one perfect string, including one deliberately wide and one deliberately narrow. Wide queries show you the shape of the market; narrow ones show whether your specific combination exists at all. Ask it to include exclusion terms too — this is what keeps agency reposts and unrelated seniority out of your results.

If you skip it: you search one title on one board and conclude there are no jobs. There usually are; they are filed under a name you did not try.

While you are here: our own open roles page is a curated feed rather than an aggregator, which makes it a useful sanity check on what a real posting for your target level looks like right now.

Step 4. Score every posting against a rubric

What you do: before reading a single posting, write the rubric. Then have the model score each posting against it and explain each score in one sentence.

Why the rubric comes before the postings: if you write it afterwards, you will unconsciously write it to justify the role you already liked. Deciding the criteria while you are calm is the only way to get an honest filter.

A workable rubric has four or five dimensions: how much of the core requirement you can evidence today; seniority match; domain familiarity; the constraints from your brief as a pass or fail rather than a score; and how clearly the posting is written, which correlates more than people expect with how clearly the team thinks.

Then sort into three piles. Apply now: you meet the core and the constraints pass. Stretch: you meet most of it and the gap is one specific thing — worth applying to a few, because postings overstate requirements. Skip: a constraint fails, or the gap is structural. Skip means skip. The pile exists so you stop revisiting it.

Ask for the sentence, not just the number

A score on its own is unusable — you cannot tell whether the model understood the posting. One sentence of justification per dimension makes the mistakes visible immediately, and mistakes at this stage are common.

Stuck on the rubric?

The fastest way to calibrate it is to score three real postings with someone who hires for those roles. That is a single conversation, and it is what the free intro call is for. If you would rather look at live postings first, the open roles page also has a weekly digest you can subscribe to.

Step 5. Read what the posting is not saying

What you do: for anything in your apply pile, ask the model what the posting implies about the team, and what it would want to ask the hiring manager.

Why: job descriptions leak information. A role that lists four distinct disciplines is usually backfilling several people with one hire. A long list of technologies with no mention of outcomes often means nobody has decided what the role is for. "Fast-paced environment" and "wear many hats" carry meanings everyone knows and nobody writes down.

What this buys you: two things. You drop roles that would have wasted four interview rounds, and you walk into the ones you keep with three sharp questions — which is, incidentally, one of the strongest signals a candidate can send.

Step 6. Verify before you spend anything

What you do: check that the role exists, on the company's own site, before you write a word for it.

Why this is not optional: models invent things, and job search is unusually exposed to it. A model asked for "companies hiring Staff Engineers in Berlin" will produce a confident list, and some of it will be wrong — companies that are not hiring, roles closed months ago, requirements it filled in from the general shape of such roles. Aggregators make it worse by keeping dead postings alive.

Check these every time

•The posting exists on the company's own careers page
•The date, and whether it has simply been reposted
•The requirements as written there, not as summarised to you
•The location and work-authorisation reality, in their words

Signs you are reading an invention

•A link that does not resolve, or resolves to a search page
•Suspiciously round salary figures with no source
•Requirements phrased more neatly than any real posting
•A company description that could describe any company

Step 7. Keep a pipeline, and let it correct you

What you do: one table. Role, company, source, date applied, score, pile, response, outcome. Anything works — a spreadsheet is fine.

Why it matters more than it looks: without it you cannot tell the difference between a resume problem and a targeting problem, and the two need opposite fixes. If well-scored applications get no reply, the resume is not landing — that is part two. If you are getting replies but from roles you do not want, the brief is wrong. If nothing scores above the line at all, your role titles are wrong and you go back to step one.

Review it weekly, not daily. Response times run one to three weeks; daily checking produces anxiety and no information. Once a week, look at what came back and change exactly one thing.

A useful benchmark

Twenty well-targeted applications with no first response is a signal worth acting on, not bad luck. Two hundred untargeted ones with no response tell you nothing at all — which is precisely why volume feels so demoralising.

The prompts

These are starting points, deliberately plain. Paste your resume or the posting alongside each one. The instruction to be sceptical is not decoration — without it, models default to encouragement, which is the opposite of useful here.

1. Extract your signal

"Here is my resume. Describe how a sceptical recruiter would read it, not how I would like it read. Give me: the role titles my experience maps to, the seniority band and why, the domains I can claim without stretching, which technologies carry real weight versus a single mention, and the three things a reader would still be unsure about."

2. Generate queries

"Here is my search brief. For each of the following boards, give me three search strings: one wide, one narrow, one focused on alternative titles for the same work. Include exclusion terms to filter out agency reposts and the wrong seniority. Explain what each string is trying to catch."

3. Score a posting

"Here is my brief, my resume, and a job posting. Score the posting from 1 to 5 on: evidenced core requirements, seniority match, domain familiarity, and clarity of writing. Mark my hard constraints as pass or fail. Give one sentence of justification per line. Then classify it as apply, stretch, or skip, and say what would have to be true to move it up a category."

4. Read between the lines

"Here is a job posting. What does it imply about the team, the scope and the reason this role is open? What is conspicuously missing? Give me three questions for the hiring manager whose answers would change whether I take this role."

Common questions

Can AI find a job for me?

No, and treating it that way is where most disappointment comes from. AI is good at reading job descriptions, naming the roles your experience maps to, generating search queries and scoring postings consistently. Deciding what you want, judging a company and verifying that a posting is real remain yours.

Why do I get no response after hundreds of applications?

Usually because the list was wrong, not the resume. Volume makes tailoring impossible, produces silence rather than feedback, and consumes the attention that the few well-matched roles needed. Twenty well-targeted applications with no first response is a signal worth acting on; two hundred untargeted ones tell you nothing.

How do I know whether a job posting is worth applying to?

Score it against a rubric you wrote before reading any postings: how much of the core requirement you can evidence today, seniority match, domain familiarity, your hard constraints as pass or fail, and how clearly the posting is written. Sort results into apply, stretch and skip.

Does AI make up job postings?

Yes. Asked for companies hiring a given role, models produce confident lists containing companies that are not hiring, roles closed months ago, and requirements invented from the general shape of such roles. Always confirm the posting on the company's own careers page before spending time on it.

How long does this method take?

About an evening to set up the brief, the queries and the rubric, then roughly an hour a week to run the search and review the pipeline.

Where this is going

Done by hand, this method takes an evening to set up and about an hour a week to run. That is a real cost, and it is the reason most people skip straight to applying.

So we are building the short version into mentors.coach: you upload your resume and get back the roles that match it — steps one through four of this guide, run for you against the same vacancy feed you can already browse. The signal extraction, the queries and the scoring happen behind the page; what you see is a ranked list with the reason for each score, so you can disagree with it.

The parts that stay yours stay yours: what you want, what you will not accept, and whether a company is somewhere you want to spend two years. If you want to be told when it ships, join the waitlist.

Until then, the manual version works. It is what the method was tested as. And if the sticking point is the resume itself rather than the search, that is where part two picks up — or you can have it done with you on a resume rebuild, or looked at end to end in a strategy deep-dive.

How to Find Jobs That Match Your Resume with AI