How to Tailor a Job Application Without Lying

Anyone applying this week can tailor a job application without lying in 25 minutes: map the posting to proof you already have, mark gaps UNKNOWN, rewrite only true lines. Run it in ChatGPT, Claude, or Gemini before you let a model invent a title, a metric, or a tool you have not touched.

Tailoring is selection, not autobiography. Show the overlap and stay quiet on the rest. A helpful model is how "helped the launch" becomes "led a $2M launch."

Read what a skill prompt actually is, then lock the proof rule below.

How to tailor a job application without lying

Proof first. Overlap second. Prose last. Start with prose and every gap gets a costume.

1. Paste the posting, then extract asks

Paste the job text. Instruct ChatGPT, Claude, or Gemini: "List the asks as bullets: skills, years, domains, tools, and outcomes. Quote a short phrase from the posting for each ask. Do not rewrite the company. Do not draft a resume yet."

If you cannot point at the phrase, the ask is the model’s idea of a job. The same quote-or-refuse habit that lets you study a PDF with AI without fake quotes applies. No phrase, no ask.

2. Map each ask to proof you can open

Now paste your real material: current resume, a few bullets from memory, links to work you can show. Instruct: "For each ask, pair one proof line I already wrote or a link I pasted. If I have no proof, write UNKNOWN. Do not infer. Do not upgrade a verb. Helped is not led."

UNKNOWN shows what to leave off, not what to hallucinate.

If your source pile is a ramble, shape it first the way you turn a messy note into a one-page brief.

3. Rewrite only the true overlap

Ask for two artifacts, nothing else:

• Resume — one page. Only lines with proof. Same dates you already use. Same titles unless you paste a legal title change.
• Note — 120 to 180 words. One reason you want this posting, two proof lines, one honest limit or UNKNOWN skill you will not fake.

Banned moves: invented metrics, rounded-up headcount, tools you have not used, "passionate about" as a substitute for proof, a company mission paragraph you could not say at dinner.

Tell the model: "If a line is not backed by my proof map, delete it. If you want a number I did not give, write UNKNOWN and stop. Do not fill."

4. Generate one variant, not a new career

Ask for a second resume only if the first dropped a proof line you care about. Do not ask for a "bolder" version. Bolder is how verbs inflate.

Check every verb against the map. Cut to one page. Do not blend resumes.

5. One human pass with the posting closed, then open

Read the resume out loud with the posting hidden. If a line sounds like the posting, not your week, cut it.

Then open the posting and check three traps:

• A tool you listed that is not on the proof map.
• A metric you cannot source in five minutes.
• A note that claims you already do the job you are applying to.

Fix those yourself. Then stop. Culture fit is a conversation, not a third paragraph. Save the proof-map loop with the Promptcrates tools for reusable writing skills.

Reusable skill card

Trigger: You have a posting you want, work you can defend, and you need a packet today without inventing a career.

Input: The posting, your current resume or proof bullets, links you can show, plus a ban on new metrics, new titles, and new tools.

Output: An ask list quoted from the posting, a proof map with UNKNOWN gaps, one one-page resume, and one 120–180 word note.

Stop: After one read-aloud pass and a source check on every number. No bolder rewrite. No blended resume. No culture paragraph.

Takeaways

• Quote the posting’s asks. No phrase, no ask.
• Map proof or write UNKNOWN. The model does not get to close gaps.
• Rewrite only true overlap. Helped is not led.
• ChatGPT, Claude, and Gemini all need a human proof pass.