Evidence-Grounded Cover Letter Prompt

Start here

This reusable prompt helps you draft a cover letter using two inputs you paste into an ordinary AI chat that accepts text: (1) the full job posting text, and (2) your own raw experience notes. Nothing is sent or applied for you.

1. Copy the complete prompt (appended with this asset).
2. Paste it into the chat, then paste the job posting and your experience notes below it (or where the prompt asks).
3. Add optional preferences if you have them: tone, length limit, things to emphasize or avoid.
4. Read the output in order: extraction, requirement map, draft, then gap and check report.

Check the output before using it: the role title and company name match the posting exactly; every number and tool name comes only from your notes; no requirement is claimed without a matching note; unsourced lines are marked for you to fix.

source reviewed; runtime not tested

Introduction

Writing a tailored cover letter usually means re-reading the posting, hunting for matching experience, and resisting the urge to overstate. This prompt structures that work: it extracts the posting's own wording, maps your notes to each requirement, drafts plain prose, then reports gaps instead of inventing achievements.

Prerequisites and permissions

• A text-accepting AI chat; no terminal or API setup is needed.
• The full job posting text and your own experience notes.
• You control what you paste. Avoid sharing sensitive personal data such as ID numbers, home address, or private contact details.
• The prompt only produces a draft. It cannot apply, submit, email, or contact anyone. Keep final review and sending in your own hands.

Limitations

• Output quality depends entirely on what you supply; missing inputs should prompt the tool to stop and ask.
• Strength labels (strong / partial / none) are the tool's judgement of your pasted notes, not verified facts.
• The number and tool rules mean a note like "cleaned up data" should trigger a request for detail, not a guessed figure.
• The worked example in the prompt is labelled fictional; it is not a tested result.

FAQ

What if I only paste the posting? The prompt should stop and ask for experience notes rather than fill in experience for you.

Can I use the draft as-is? No. Treat it as a starting point; confirm every claim against your own records and fill in all placeholders before sending.

Attribution

Original TokRepo prompt, licensed CC BY 4.0. External reference material retains its own rights. Reference: ChatGPT release notes. The complete prompt and translation are appended with this asset.

Complete reusable prompt

You are helping me draft a cover letter. You write from the material I supply; you do not add facts about me, the employer, or the role that are not in my input.

INPUTS I WILL PASTE

1. JOB POSTING — the full text of the listing (responsibilities, requirements, any stated values or process notes).
2. MY EXPERIENCE NOTES — raw bullets about roles, projects, skills, results, tools, and motivation. Include numbers only if I give them.
3. OPTIONAL — my preferred tone (default: warm-professional), length limit, and anything I want emphasized or avoided.

If the posting or my experience notes are missing, stop and ask for them. If a section is unreadable or truncated, say so and ask me to re-paste rather than guessing.

STEP 1 — EXTRACT (show me this first) From the posting, list: the role title and company name exactly as written; the top 5–7 requirements or responsibilities in the posting's own phrasing; and any named tools, methods, or keywords. Do not paraphrase requirements into different claims.

From my notes, list: which of those requirements each item plausibly supports, and which requirements have NO supporting item. Flag any bullet that is vague (e.g. 'improved processes') and ask me for one concrete detail I could truthfully add.

STEP 2 — MAP Produce a short table: Requirement | My supporting evidence | Strength (strong / partial / none). For 'none' rows, do not invent experience. Instead propose a neutral line I could use to show adjacent capability or willingness to learn, clearly labelled as an option for me to accept or reject.

STEP 3 — DRAFT Write the letter in plain, specific prose:

• Opening: the exact role title, and the single strongest true reason I am a fit drawn from my notes.
• Body: 2–3 short paragraphs, each anchored to a named requirement from the posting and one item from my notes. Where I gave a number, use it exactly; never round up or extrapolate.
• Close: a brief, non-presumptuous line about next steps, plus my name placeholder.
• Do not use clichés like 'I am writing to express my strong interest' unless I ask. No superlatives I cannot support ('world-class', 'unmatched').
• Keep to my stated length, or 250–350 words by default. One page.

STEP 4 — GAP + CHECK REPORT After the draft, list: (a) Requirements with no evidence, and every placeholder I must fill. (b) Every place a reader could ask 'how do you know that?' — quote the line and the source bullet it came from, or mark it UNSOURCED for me to fix. (c) Three questions a hiring manager might ask that my draft does not yet answer. (d) A final self-check: does the letter name the correct role and company; is every claim traceable to my notes; is it within the length target; does it avoid restating the whole résumé?

BOUNDARIES You are preparing a draft only. You cannot apply, submit, email, or contact anyone. Do not promise outcomes, invent credentials, dates, employers, or metrics, or claim I have skills I did not list. If my notes conflict with each other, flag the conflict instead of choosing silently.

WORKED EXAMPLE (fictional) Input posting line: 'Support weekly reporting and improve data quality in our CRM.' My note: 'Cleaned 400 duplicate client records in Salesforce over two weeks.' Illustrative requirement-map row: Support data quality | 400 duplicate records cleaned in Salesforce in two weeks | strong Illustrative draft sentence: 'In my last role I cleaned 400 duplicate client records in Salesforce over two weeks, which is the kind of data-quality work your weekly reporting depends on.' Pass/fail checks on that example: the number 400 and the tool name come only from my note (pass if unchanged; fail if inflated or tool invented). The posting's word 'CRM' is echoed, not replaced by a different system. If my note had said only 'cleaned up data', the correct output would be a request for the record count, not a guessed figure.

References and reuse

• ChatGPT release notes · Reviewed 2026-10-04

Original TokRepo prompt · CC BY 4.0. Reference documents retain their own rights.