Short answer: Quick answer: turn the job description into an input pack: must-have requirements, desired outcomes, company context, and collaboration signals. Map only the strongest requirements to evidence you can prove, ask questions for missing proof, then audit generic or unsupported lines before sending.
Applicants who already have a draft or proof notes and need to adapt the letter to a specific job description, job offer, or role posting.
Avoid using it to copy every bullet, invent employer research, hide missing proof, or stuff the letter with keywords from the posting.
Paste the job offer, build a signal-to-proof table, answer missing-proof questions, then rewrite only the paragraphs that do not match the role.
Build an input pack from the job offer
Do not paste the offer and ask for prose immediately. First split the job description into must-have requirements, desired outcomes, company context, collaboration signals, and tone clues.
Pick the top three signals
Do not answer every bullet. A good letter answers one or two signals deeply and leaves the rest to the resume. Mark any strong signal with no proof as a gap.
Map each signal to proof before rewriting
For every signal, write one resume proof point, source, and confidence level. If you cannot prove it, ask a follow-up question instead of letting AI pretend you can.
Requirement → proof → boundary: three worked examples
Use each example as a decision record, not a sentence to copy. Name the job requirement, point to the evidence source, and write the boundary that keeps the claim interview-defensible.
- Cross-team collaboration → proof: weekly handoff notes and two resolved dependencies from a named project → boundary: say you coordinated those handoffs, not that you owned the whole program or every stakeholder decision.
- Experience with a required tool → proof: a dated dashboard, repository, certification, or course artifact → boundary: label a sandbox, course, or personal project honestly; do not turn practice into production ownership.
- Improve a measurable outcome → proof: a report with the baseline, date range, sample, and your exact action → boundary: use only sourced numbers and do not claim sole causality for a team result.
Prompt
Send-before verification: requirements, evidence, judgment
Before you attach the letter, scan every paragraph through this three-part check. Each sentence should survive the requirement test, the evidence test, and the human-decision test.
- Requirement → paragraph match: Name the job requirement each paragraph answers. If a paragraph serves no requirement from the job offer, delete it or merge its proof into another paragraph.
- Missed requirement check: Scan the job offer again. If a must-have requirement matches your experience but you did not address it, add one sentence with a named source.
- Evidence source: For every factual claim, write down where the proof lives — resume bullet, performance review, project doc, certification, or reference. Claims without a source are unsupported.
- Number audit: Flag any metric, percentage, or dollar amount you cannot pull from a document. Remove AI-invented numbers even if they sound reasonable.
- Decision: Keep — The claim answers a job requirement, traces to a named source, and you can defend it in an interview without pausing.
- Decision: Rewrite — The claim targets the right requirement but uses vague or AI-polished language. Replace with one concrete detail from your source.
- Decision: Delete — The claim has no source, answers no requirement, or fits any applicant at any company. Also cut invented enthusiasm, unverified company praise, and unsupported metrics.
FAQ
Can AI adapt the same cover letter to many jobs?
It can adapt structure, but each job offer needs fresh proof mapping. Reusing the same evidence without checking the role signal makes the letter interchangeable.
What should I check before sending?
Check that every claim has resume proof, the company reason is verified, missing requirements are not hidden, and no paragraph could fit any other job unchanged.
What if my only proof comes from a course, personal project, or adjacent role?
Use it if it answers the requirement, but name the source and artifact precisely. A course dashboard, personal repository, or adjacent-role handoff can show relevant ability; it does not prove production ownership, a formal title, or a result you cannot source.