Job search: CVs and cover letters with AI

Beginner · 6 min · Tailor a CV or cover letter with AI without exaggerating.

job search
writing

You can use AI to tailor a cover letter or CV to a specific posting — as long as every claim stays traceable to something you actually did. The workflow has four moves: gather evidence, map it to the posting, draft inside the truth, and check. Skip the first move and the model fills the gaps with plausible fiction you will have to defend in an interview.

Gather evidence before anything else

Write a one-page brag sheet: roles and dates, projects and what you personally did, numbers you can defend, tools you actually used, outcomes with context. This is your grounding bank — the only material the model is allowed to draw from. It does not know your history, and an ungrounded draft is a nicer-sounding guess about a stranger.

Map the posting

Turn the posting into a requirements list, then mark each line:

Evidence. Covered, with the specific proof.
Partial. You touched this. Say how, at the true scale: “familiar with X from Y”, not “expert in X”.
Miss. Not covered. Either leave it alone or address it honestly in one line — never let a bullet silently absorb it.

This map is also your cover letter’s outline: requirements on the left, evidence on the right, and the letter argues the strongest three or four pairings.

Draft inside the truth

Weak prompt

Take my CV and make it sound really impressive for this marketing manager job. Don't hold back — make my experience sound more senior than it is written, and add anything that would help me stand out.

Stronger prompt

Here is the posting's requirement list and my evidence bank. Map each requirement to a bullet: the closest evidence, or “gap”. Then draft a 210-word letter using only the evidence bank — same facts, same scale, nothing upgraded. If a requirement has no evidence, ask me for a fact instead of inventing one. End with the three strongest requirement-evidence pairings as a bullet list.

Why the stronger one works

“Sound more senior” is an instruction to fabricate; the model complies happily and you own the claim in the interview.
The strong prompt supplies the two artefacts that make honesty easy: the requirement list and the evidence bank.
“Ask me for a fact instead of inventing one” gives the model a legitimate path when evidence is missing.

The check is reading your own draft

Before sending, read every sentence and point to the evidence behind it. Anything you cannot point to is an unsupported claim — cut it or correct it to the true scale. Two failure patterns to hunt: scope upgrades (“led” when you contributed; “owned” when you supported) and time stretching (“years of” where the evidence is a project). Both read fine and both are easy for an interviewer to puncture with one follow-up question.

Your tailored draft claims you “led the migration of the billing system”. In reality you were one of six engineers on it. What do you do?

Keep it — it reads better, and the team did migrate the system Reword to the true scope — “contributed to the migration of the billing system as part of a six-engineer team” — then check the rest of the draft for similar upgrades Add “team player” elsewhere in the letter to balance it out

Practice

Fix the cover letter prompt

A friend is applying for a product analyst role and asks for help with the AI prompt they have been using. Their last two letters read well in the moment, but in one interview they were asked to explain a claim they could not back up.

Rewrite the prompt so the model tailors the letter to the posting using only real evidence — and asks for missing facts instead of inventing them.

The weak prompt

Take my CV and write a cover letter for this product analyst job. Make it really impressive and confident — add anything that would make me stand out, and make my experience sound more senior than my CV shows. Don't worry about length.

Your improved version

Self-check — does your version address these?

Sets a truth boundary: no facts beyond the supplied evidence — Say explicitly that the model may not add, upgrade or stretch anything — and what to do instead. Feeds the posting's requirements as a list to map against — A cover letter is an argument about fit; the requirement list is the argument's skeleton. Supplies the evidence bank (or asks for it) and requires gaps to be labelled — Roles, projects, numbers and outcomes the person can defend — plus the instruction to mark what is missing. Constrains the output for the actual reader — A length cap, plain structure, and the strongest requirement-evidence pairings made visible. Adds a self-check: every sentence traceable to the evidence bank — Ask the model to flag any sentence it cannot trace — that flag is the interview-proofing step.

Hint

Replace every invitation to embellish (“impressive”, “more senior”) with an instruction about fidelity.

The model needs two artefacts to be honest: the requirements list and the evidence bank — name both.

Transfer

Interviews: the evidence bank doubles as your prep sheet — every claim has a story behind it.
Performance reviews: same discipline, internal audience; upgrades age badly when promotions are decided.
Freelance profiles and bios: map claims to deliverables, cut anything you cannot show.
LinkedIn summaries: write from the evidence bank, not from the job title you wish you had.

Key takeaways

Evidence first: an ungrounded draft is fiction with good formatting.
Map every posting requirement to evidence, partial coverage, or an honest gap.
Check your own draft line by line — scope upgrades and time stretching are the failures to hunt.

Keep going on the practice floor, or check where you stand with the skills check.
Related: Find the unsupported claim is the same check applied to other people’s text.
Terms: grounding, unsupported claim and caveat.