How to use ChatGPT to write a cover letter (without the AI red flags)
Can employers tell you used ChatGPT? What OpenAI's own detector got wrong, what to strip before you paste your resume, the prompt, and the edit that fixes the draft.
ChatGPT will write you a cover letter in ten seconds. The letter will be grammatical, polite and three paragraphs long, and a recruiter who has read forty of them this week will recognize it before the second sentence. Detection software is unreliable at flagging it. The letter is about nobody in particular.
That is fixable, and most of the fix happens before you open ChatGPT. This guide covers what employers can and cannot detect, the order of work one AI company asks its own candidates to follow, what to remove from your resume before you paste it anywhere, the prompt, and the ten-minute edit that turns the draft into your letter.
Can employers tell you used ChatGPT?
Not with software, at least not reliably. The best evidence comes from OpenAI itself. In January 2023 it released a classifier meant to tell AI-written text from human text, and published its accuracy:
"In our evaluations on a 'challenge set' of English texts, our classifier correctly identifies 26% of AI-written text (true positives) as 'likely AI-written,' while incorrectly labeling human-written text as AI-written 9% of the time (false positives)."
The same page warned that "the classifier is very unreliable on short texts (below 1,000 characters)." On July 20, 2023, OpenAI withdrew it, writing that "the AI classifier is no longer available due to its low rate of accuracy" (OpenAI).
26%
and 9% of human text wrongly flagged
of AI-written text caught by OpenAI's own detector before OpenAI withdrew it for low accuracy. A cover letter is short text, the case the company called "very unreliable".
Source: OpenAI, New AI classifier for indicating AI-written text, 2023 (update of July 20, 2023)
A cover letter is exactly the kind of text detectors struggle with. HiredFolio's own letter generator, for comparison, aims at 900 to 1,300 characters for a short letter and 1,300 to 2,200 for a medium one, so a short letter sits right around the 1,000-character line OpenAI drew.
So the real answer to "can they tell" is that they read. A recruiter doesn't need a detector to notice that your letter could be signed by any other applicant. The red flags further down are what they notice, and every one of them can be removed by hand.
Write the first draft yourself, then let ChatGPT refine it
Every guide we read in the top search results for this topic teaches the same order: paste the job post, ask ChatGPT for a letter, edit what comes back. At least one employer publicly asks for the opposite.
Anthropic, which makes the Claude assistant, publishes a page called "How to collaborate with Claude during our hiring process" (last updated July 10, 2025). Its instruction for the application stage reads:
"Please create your first draft yourself, then use Claude to refine it. We want to see your real experience, but Claude can polish how you communicate about your work."
The page lists "Using Claude to better articulate your experience" as encouraged and "Using Claude to create experiences" as not allowed (Anthropic). On October 5, 2026, every job posting on Anthropic's public Greenhouse board linked to that policy.
You don't need to apply there for the order to be useful. A draft you wrote yourself, however rough, already contains the facts only you know: the client you kept, the backlog you cleared, the reason you want this company. ChatGPT is very good at tightening sentences. It cannot add a fact it was never given. When it starts from a blank page, it fills the gaps with adjectives.
A rough draft can be five bullet points:
- the job title and one thing about the company that made you click Apply;
- your two results closest to the posting, with numbers;
- how you got each result, in one line;
- one requirement from the posting you match exactly;
- when you can start, or what you'd like to talk about.
Then use the prompt below to turn the bullets into a letter.
What to paste into ChatGPT, and what to strip first
Most guides tell you to paste your whole resume. Before you do, know where it goes. OpenAI's help center says, for ChatGPT and its other individual plans: "we may use your content to train our models." You can turn this off: "To opt out, turn off Improve the model for everyone under Settings > Data controls in ChatGPT." And: "Temporary chats are not used to improve OpenAI models while they remain temporary" (OpenAI Help Center, checked October 5, 2026).
Three habits follow from that:
- Strip your contact block. Phone number, email, street address and LinkedIn URL add nothing to the letter's argument. Type the header yourself at the end.
- Use a Temporary Chat or turn off training before pasting anything about a current employer, especially client names or revenue figures.
- Keep confidential numbers relative. "Cut onboarding time in half" makes the same point as the exact contract value of your biggest account.
For the record, HiredFolio's generator (which calls OpenAI's API) sends the model your name, resume title, summary, selected skills and experience lines. Your email and location are not part of the prompt. OpenAI's stated default for its API is: "By default, we don't use inputs or outputs from ChatGPT Business, ChatGPT Enterprise, ChatGPT Edu, or our API to improve our models."
The prompt that gives you a usable draft
Paste the whole job posting, not just the title. The posting is where the model finds the tools, missions and vocabulary the letter should echo. Then paste your bullet draft and the resume experience lines closest to the role.
You are a hiring manager who has read five hundred cover letters.
Turn my notes below into a cover letter for the job posting below.
Rules:
- Three paragraphs, 250 words maximum. No "I am writing to apply".
- The first sentence is about the company or its customers, not about me.
- Every claim comes from my notes or my resume. Do not invent any number,
employer, tool or responsibility. If something is missing, write [ADD]
instead of guessing.
- First-person action verbs ("I cut", not "I helped to reduce").
- Banned: passionate, dynamic, results-driven, team player, "I am confident
that", "perfect fit", "leverage", "excited to apply".
- Use the posting's exact words for tools and responsibilities.
- End with one sentence proposing a conversation.
Job posting:
[paste the full posting]
My notes:
[paste your five bullets]
Resume, relevant experience only:
[paste two or three roles, without contact details]
The [ADD] line matters most. It gives the model an honest way out when it lacks a fact, instead of inventing one. If the draft comes back with three [ADD] markers, those are the three facts your letter was missing.
Prompt add-ons for your situation
Add one line to the rules above.
- No experience or first job. "I have no full-time experience. Use my coursework, internship and volunteer work as evidence, and name the specific project for each claim." Our guide to a resume with no experience covers which school projects count.
- Career changer. "I am moving from [old field] to [new field]. Open with the reason for the switch in one sentence, then map two results from my old job onto two requirements of this posting." The career change resume guide has the same mapping for the resume side.
- Mid-career. "Pick the two results from my notes that match the posting's top two requirements, and leave the rest out."
- Senior or management. "Lead with the scale I managed (team size, budget, accounts) and one decision I made, not a list of responsibilities."
- Employment gap. "Mention the gap in one neutral sentence in the last paragraph, with what I did during it. Do not apologize." Our page on resume gaps has wording that holds up in an interview.
The red flags recruiters actually notice
None of these are unique to AI. People wrote them for decades. AI just produces all of them at once.
- A letter that fits any company. The company's name is the only specific detail. Fix: one sentence about their customers, product or a line from the posting, tied to one fact about you.
- Adjectives instead of evidence. "Detail-oriented and results-driven" with no result. Fix: replace each adjective with the result that would have earned it.
- Invented facts. When the prompt lacks information, the model fills the gap, and a requirement from the posting can come back as a skill you claim. Read every claim against your resume. Anything you can't back up in an interview goes.
- The posting copied back. Echoing the posting's tool names helps. Pasting back its sentences ("you are seeking a self-starter who thrives in a fast-paced environment") reads as a mirror, not a match.
- Three paragraphs of identical shape. Each opens with a quality and closes with "which allows me to". Vary the length. Let one paragraph be two sentences.
- The weak verb. "Contributed to", "was involved in", "helped". If you did it, say so.
Example a typical generated sentence, rewritten with real facts
Generated
With my strong organizational skills and passion for operational excellence, I am confident I would be a valuable asset to your logistics team.
Edited
At my current warehouse I rebuilt the pick-list schedule so the late shift stopped waiting on morning returns, and same-day dispatch went from 82% to 95% of orders in four months.
The ten-minute edit
Once the draft is back, work through it in this order.
- Check every fact (3 min). Highlight each number, employer, tool and responsibility. Each one must appear in your notes or resume. Delete anything that doesn't.
- Rewrite the first sentence (2 min). Make it about the company, and connect it to one thing you've done. If ChatGPT's opener is about you, move it down or cut it.
- Add the "how" to each result (3 min). In the example above, "dispatch went from 82% to 95%" is a claim. Adding "by rebuilding the pick-list schedule" makes it evidence an interviewer can ask about.
- Run the signature test (2 min). Read each sentence and ask whether another applicant could sign it. If yes, rewrite it with a fact or delete it.
Read it aloud once at the end. Any sentence you wouldn't say to the hiring manager in person gets rewritten. The format and the parts of the letter itself are covered in our guide on how to write a cover letter.
ChatGPT and your cover letter what shows and what holds up
Shows right away
A one-line prompt with only the job title
Your full resume pasted, phone and address included
A number or skill the model guessed
The same letter for fifteen postings
Holds up in the interview
Your own five bullets first, then the prompt
Contact details stripped, training turned off
Every claim checked against your resume
One prompt per posting, the full posting inside
ChatGPT writes well and knows nothing about you. Every sentence it writes without your facts is a sentence anyone could send.
Generate the letter from your resume with HiredFolio
The prompt above needs your notes, your resume and the full posting pasted in for every application. HiredFolio does that step for you. You save the job from the posting, pick a resume, and the letter is generated from both in one AI action. The prompt behind it tells the model: "Do not invent experience, employers, degrees, certifications, or metrics." You choose the length and tone, and the letter stays attached to that application along with the resume you sent.
It doesn't know the company beyond the posting, and it can't write the sentence that only you can, about why this company. The ten-minute edit still applies. What it removes is the copy-pasting and the blank page.
Generate a cover letter from your resume with HiredFolio
FAQ
Should I use ChatGPT to write my cover letter?
Yes, as an editor. Write the facts yourself, let ChatGPT shape them, then check every claim. Using it to invent experience is the line both recruiters and published employer policies draw.
Can ChatGPT write a good cover letter on its own?
It writes a well-structured one. It can't write a specific one, because the specifics (your results, why this company) aren't in its training data. They're in your head.
Do I need to edit a ChatGPT cover letter?
Always. At minimum, check every fact against your resume and rewrite the first sentence. That's the part a recruiter reads before deciding to read the rest.
Is it okay to use ChatGPT for job applications?
Usually, unless the posting says otherwise, so read the application instructions. Some employers publish a policy. Anthropic's, for example, encourages AI for refining your draft and bans it for creating experiences or for take-home assessments unless allowed.
How long should a ChatGPT cover letter be?
Short enough to read in one pass. Ask for a word limit in the prompt (250 words fits on one page with room to spare), because without one the model pads.
Key takeaways before your next application
Write five bullets of your own facts before you open ChatGPT, then ask it to shape them, not invent them.
Remove your contact details and turn off "Improve the model for everyone" or use a Temporary Chat before you paste your resume.
Check every number and skill in the draft against your resume, then run the signature test on each sentence.
The letter is only as good as the resume behind it: make sure yours states its results.
Read next
- Letter of interest: samples, a template, and what the data says about sending one
How to write a letter of interest to a company that isn't hiring: a template, five samples, the letter of intent difference, and survey data on when it works.
- How to end a cover letter (with examples)
How to end a cover letter: the last sentence that earns a reply, which sign-off to use in the US and UK, how to sign a PDF or an email, and what 31 university sample letters actually do.
- How to write a motivation letter: what the form actually asks for, with examples
How to write a motivation letter for a master's, a scholarship, an internship or a job: the limits real forms impose, and a full annotated example.