Quick answer
To write a cover letter using ChatGPT, give it the job ad, the company's own words and your resume, then demand four paragraphs under 320 words where each one carries a specific proof. Run a second prompt that removes AI vocabulary and checks every claim against your resume.
Writing a cover letter using ChatGPT takes about four minutes. Writing one that survives a hiring manager’s first ten seconds takes a little longer — and almost none of the extra work is what the usual guides tell you to do.
The standard advice is: paste your resume, paste the job ad, ask for a cover letter. That works, in the sense that words appear. It also produces roughly the letter several hundred other applicants are sending to the same role this week. Here’s what actually matters — hiring managers rarely reject a letter for being written with AI. They reject letters carrying no evidence at all:
- Nothing specific about the company that took more than a glance to find.
- No numbers from your own record — only adjectives about your attitude.
- No reason the letter couldn’t have been posted to any employer in the country.
That’s a structural problem, and structure is exactly what a prompt can fix. What follows is the Four-Proof Letter: a four-paragraph structure where every paragraph carries one piece of evidence only you could supply, plus a second pass that strips out the vocabulary that makes AI writing obvious.
Will this work for you?
Why Most AI Cover Letters Fail
A cover letter using ChatGPT usually fails because it describes enthusiasm instead of proving fit. The model has your resume and the job ad, so it does the obvious thing — restates your resume in more formal language, then adds a layer of eagerness on top. Nothing in the output is checkable, and nothing is specific to the reader.
✕ The default opener
"I am writing to express my enthusiastic interest in the Operations Lead position at your dynamic organisation. With a proven track record of delivering results, I am confident I would be a valuable addition to your team."
✓ An opener with a proof in it
"Your job ad mentions three warehouse sites moving to a single inventory system by March. I ran that exact migration at Brightfold last year — four sites, eleven weeks, no stock freeze."
Recruiters skim two lines before deciding whether to read the rest. An opener that names something real — a product, a recent announcement, a problem stated in the job ad — buys you the remaining paragraphs. An opener about your own excitement buys nothing.
The tells that give it away
- Opening with “I am writing to apply for” or any variant of being thrilled, excited or delighted.
- Vocabulary nobody uses out loud: tapestry, testament, spearheaded, best-in-class.
- Adjectives where numbers belong — “significantly improved” instead of “cut it from 9 days to 4”.
- Praise for the company that would be true of any competitor in the same sector.
- Four hundred and fifty words of it, when three hundred would have done more.
You don’t fix these by asking for something “more human”. You keep them out of the draft in the first place, then run a pass that removes whatever slipped through.
The Four-Proof Letter, Explained
The Four-Proof Letter is a cover letter structure where each of the four paragraphs carries exactly one piece of evidence nobody else could have written for you. Paragraph one proves you looked at them. Paragraph two proves you can do the work. Paragraph three proves the first two connect. Paragraph four makes saying yes easy.
That constraint is the entire method — the Four-Proof Letter approach we teach at PromptPeakAI — and it works because it gives the model something specific it can fail. “Write a compelling cover letter” has no failure condition, so anything counts as success. “Paragraph two must contain two numbers from my resume and no adjectives about my work ethic” can be checked in five seconds.
Four paragraphs, four proofs. Every paragraph has to earn its place with something checkable.
| Paragraph | The one proof it carries | Words |
|---|---|---|
| 1. The Anchor | A specific detail about the employer | 50 to 70 |
| 2. The Evidence | Two measurable results from your record | 80 to 100 |
| 3. The Bridge | The link between their problem and your record | 70 to 90 |
| 4. The Ask | A concrete, easy next step | 30 to 40 |
Four paragraphs, roughly 280 to 320 words in total. Longer and you’re writing a second resume. If you already use the Role, Goal, Context, Format structure, a cover letter using ChatGPT is simply that structure with an unusually strict Format specification.
Protect Your Resume Before You Paste
Sort your privacy out before you write a cover letter using ChatGPT, because a resume is one of the densest personal documents you own: full name, home address, phone number, ten years of employment history, sometimes a salary figure.
Here’s the part most guides get wrong. The data-training boundary is not free versus paid. Per OpenAI’s own documentation on how your data is used, content from consumer accounts — Free, Go, Plus and Pro alike — may be used to train its models unless you opt out, while business and enterprise accounts are excluded by default. Paying for a personal subscription changes nothing here.
Job ads, public About pages, your own achievements and figures.
Home address, phone number, personal email, exact current salary.
Employer-confidential revenue, unreleased plans, anything under NDA.
Two settings worth thirty seconds
- Turn training off. Profile icon, then Settings, then Data Controls, then switch off “Improve the model for everyone”. On mobile it’s the side-bar menu, your profile icon, then Data Controls. The setting applies to your whole account on every device.
- Or use a Temporary Chat. These aren’t used for training, don’t create memories, don’t appear in your history, and are deleted after 30 days. For a one-off job application, this is the cleaner option.
The ChatGPT Settings modal open on the Data Controls tab, with the Improve the model for everyone toggle switched off.
Crop to the settings panel only. Hide the account email address, profile photo, and the entire conversation sidebar.
File: chatgpt-data-controls-improve-model-toggle.png · Alt: "ChatGPT Data Controls settings with the improve the model for everyone toggle turned off"The ChatGPT message box showing the attachment control and the Temporary Chat toggle in the top corner of the chat screen.
Crop to the chat input area and top bar only. Hide browser tabs, bookmarks, and any previous conversation titles.
File: chatgpt-temporary-chat-and-attach.png · Alt: "ChatGPT input box showing the file attachment control and the Temporary Chat option"Then redact the document itself. Swap your street address for the city, cut the phone number and personal email, and round any figure belonging to your employer rather than to you. None of it improves the letter — real contact details go in your word processor at the end. If you’re job hunting from a current role, what ChatGPT does with work data is worth ten minutes first.
Build Your Cover Letter Using ChatGPT
The workflow runs in two stages: ChatGPT drafts, then you decide. Stage one gathers real material and produces a structured draft. Stage two removes the AI vocabulary and checks every claim against your actual record.
Two stages, five steps. The model never gets the last word.
Collect the raw material
Open a Temporary Chat and give the model three things: the full job ad, a paragraph or two from the company’s About page or a recent announcement, and your redacted resume. On the Free plan, pasting the text works more reliably than attaching a file — file uploads are limited there, and a long resume plus a long job ad can bump into the input ceiling.
You are helping me research an employer before I write a cover letter. Below is a job advert and some text from the company's own website. JOB ADVERT: [paste the full advert] COMPANY TEXT: [paste the About page or a recent announcement] Give me: 1. The three specific problems this role is being hired to solve, quoted from the advert. 2. Two concrete facts about this company that a lazy applicant would not know. Use only what is in the text above. If something is not stated, say "not stated" rather than guessing.
You should now have three named problems and two specific company facts — all traceable to text you pasted.
Pull two provable numbers
This step separates a real letter from a polished one. Most resumes contain numbers that never reach the cover letter because they’re buried halfway down page two.
Here is my resume: [paste your redacted resume] And here are the three problems the employer needs solved: [paste the three problems from Prompt 1] Find the two achievements in my resume that most directly address those problems. For each one, state it as: what the situation was, what I did, and the measurable result. Rules: - Use only achievements that appear in my resume. Do not invent, round up, or embellish any number. - If my resume does not contain a measurable result for an achievement, say so instead of estimating one.
✕ What the model reaches for first
"I have a proven track record of driving significant operational improvements and leveraging cross-functional relationships to deliver best-in-class results for stakeholders."
✓ What the resume actually supports
"I cut invoice processing from nine days to four across three sites, and took the supplier query backlog from 340 open tickets to under 50 in one quarter."
Run the Four-Proof drafting prompt
Now the structure goes in. Notice how much of this prompt is telling the model what not to do — the banned list does more work here than any instruction about tone.
Write a cover letter for the [job title] role at [company]. Use exactly four paragraphs, 280 to 320 words in total: Paragraph 1 (50-70 words): Open with one specific fact about this company from the notes below. No greeting clichés. Do NOT begin with "I am writing to apply" or any statement of how excited I am. Paragraph 2 (80-100 words): My two achievements below, written as situation, action, measurable result. Numbers, not adjectives. Paragraph 3 (70-90 words): Connect those achievements to the problems this role must solve. Be explicit about the link. Paragraph 4 (30-40 words): A short, confident closing with one concrete next step. BANNED WORDS: delve, tapestry, testament, spearhead, synergy, leverage, seamless, dynamic, passionate, thrilled, excited, proven track record, best-in-class, valuable addition. OTHER RULES: - Use only facts I have given you. Invent nothing. - Plain business paragraphs. No markdown, no headers, no bracketed stand-ins except the recipient's name. - Vary sentence length. Contractions are fine. COMPANY NOTES: [paste from Prompt 1] MY ACHIEVEMENTS: [paste from Prompt 2]
You should now have a four-paragraph letter under 320 words that names the company specifically and contains at least two real numbers.
The Strip Pass That Removes Tells
The strip pass is a second prompt that reads the draft back and removes the residue — leftover formality, adjective clusters, sentences that all run to the same length. Run it even when the draft looks fine. The tells you stop noticing in your own cover letter using ChatGPT are the ones a recruiter reads fifty times a week.
Strip the AI vocabulary
Rewrite the letter below so it sounds like a confident professional talking to a colleague, not an applicant performing eagerness. Rules: - Keep every fact and every number exactly as written. Change only the language. - Remove any word from this list if it survived: delve, tapestry, testament, spearhead, synergy, leverage, seamless, dynamic, passionate, thrilled, robust, meticulous, honed. - Break up any sentence over 25 words. Mix short sentences with longer ones. - Cut every adjective that is not doing real work. - Active voice throughout. Then list, separately, any sentence you think still reads as machine-written. LETTER: [paste the draft]
That last instruction is the useful one. Asking the model to flag its own weakest sentences gives you a shortlist to rewrite by hand, which is where your actual voice enters the document. This is negative constraint prompting doing what it does best — a banned list beats any amount of describing the tone you want.
Same facts, two very different letters. The difference is one extra prompt and five minutes.
Verify every claim against your resume
Models fill structural gaps with plausible-sounding detail, which in a cover letter means invented job titles, certifications you don’t hold, and numbers that drifted upward between drafts. Run one final check before the letter goes anywhere near an application portal.
Compare the letter below against my resume, line by line. List every factual claim in the letter — job titles, dates, numbers, tools, certifications — and mark each one as either CONFIRMED (it appears in my resume) or NOT FOUND. Do not fix anything. Just produce the list. RESUME: [paste your resume] LETTER: [paste the stripped letter]
Every claim should come back CONFIRMED. Anything marked NOT FOUND gets deleted or corrected by hand — never re-prompted into existence.
This is the same discipline that applies to any AI output you put your name on, and it matters more here than almost anywhere else, because an invented certification on a cover letter is a problem you’ll be asked about in the interview.
When the Output Goes Wrong
Four failures account for nearly everything that goes wrong with this workflow. Each has a fix at the prompt level rather than the editing level.
| What you see | Why it happens | The fix |
|---|---|---|
| Invented job titles or certifications | The model filled a gap you left open | Add the honesty check from Prompt 5 |
| Markdown headers and bracketed stand-ins | Default formatting for structured text | Ask for plain business paragraphs explicitly |
| The letter runs past 450 words | Models pad when given several conditions | Set a hard word ceiling per paragraph |
| It just rephrases your resume | No target problem was named | Feed in the three problems from Prompt 1 |
If the portal caps you at 250 words, don’t cut sentences out of a 320-word letter — the narrative breaks and it reads as though a paragraph went missing. Re-run Prompt 3 with the tighter ceiling instead. It’s a good general lesson in refining AI output by changing the input.
Does the Free Plan Handle This?
Yes. Writing a cover letter using ChatGPT works fine on the free tier, and for someone applying to a handful of roles a week there’s no case for upgrading. The limits that exist are real but easy to work around.
The one worth knowing is the input ceiling. According to ChatGPT’s published plan comparison, the free tier’s everyday model takes roughly twelve pages of text in a single input, against roughly forty pages on Go and Plus. A three-page resume and a two-page job ad fit comfortably. A ten-page CV plus a long advert plus three pages of company background will not.
| What matters here | Free | Go and above |
|---|---|---|
| Every prompt in this guide | Works | Works |
| Input size per message | About 12 pages | About 40 pages |
| File uploads | Limited | Included |
| Building a reusable cover letter GPT | Not available | Available |
Paste or attach? On the free plan, the answer is usually paste.
The honest upgrade trigger is volume, not quality. At ten or more tailored applications a week, a reusable custom GPT holding your resume and banned-word list saves real time — and that needs a paid tier. Full breakdown in our guide to choosing between the free and paid tiers.
What ChatGPT Cannot Do Here
Being straight about the limits is what keeps this workflow from backfiring on you in an interview room.
Four things stay yours, and outsourcing them is how good applications go wrong:
- Knowing which achievement matters. The model weights everything in your resume equally. You know which project the interviewer will actually care about.
- Judging tone against culture. A council HR team and an eight-person startup want different letters. ChatGPT cannot read the room from a job advert.
- Owning every claim. You will be asked about the numbers in that letter. If you cannot talk for two minutes about how you got them, they should not be in there.
- Deciding whether to apply at all. A well-written letter for the wrong role is still a wasted afternoon.
One more thing worth saying plainly: no applicant tracking system rejects you for using AI. Those systems parse and rank text against the job requirements; they don’t run detection. What gets you rejected is a letter written for nobody in particular — a content failure, not a technology one. So read the final draft out loud once, and change any sentence you’d never say to a person.
After it’s sent, the same evidence carries forward. Pull your two proof points into your interview preparation, because you’ll be asked to expand on exactly those numbers, and keep them for the salary conversation later. Prompt 2 is the reusable part.
Build your Four-Proof letter today
- Open a Temporary Chat so your resume is not used for training. Redact your address, phone number and any employer-confidential figures. Extract the three problems the job ad says this role must solve. Pull two measurable achievements from your resume that answer them. Draft four paragraphs under 320 words with the banned-word list attached. Run the strip pass, then check every claim against your resume by hand.