How to Use AI to Write a Cover Letter That Actually Gets Read
AI writes a good cover letter only when you give it proof, not a job title. Paste in the job ad and a short list of your real achievements, tell it to use only those facts, and you get a draft worth editing. Paste in “write me a cover letter for a marketing job” and you get the same polite, forgettable letter every other applicant just sent. The method in this guide is the first one: six steps, copy-paste prompts, and a worked example from a real-style job ad to a finished letter.
The stakes are higher than they look, because almost everyone is doing this now. Gartner’s 2025 survey of 3,290 job candidates found that 39% used AI during the application process, and half of those used it to generate cover-letter text. Greenhouse’s 2025 AI in Hiring Report, which surveyed 4,136 job seekers, recruiters, and hiring managers, found that 74% of US job seekers personally use AI — and that 65% of hiring managers have caught applicants using AI deceptively, from reading AI-generated interview scripts to hiding prompt injections in resumes. (Greenhouse, 2025 AI in Hiring Report) The letter that gets read is not the one that hides its AI help best. It is the one where every sentence could only have been written by you, about this job.
What Should I Give the AI Before It Writes Anything?
Two things, in this order: the job ad, decoded, and your proof. Skip either one and the AI fills the gaps with invented achievements and generic praise — the two things hiring managers notice first.
Start by asking the AI to read the job ad like the hiring manager who wrote it. Paste the full ad and use this prompt:
“Here is a job ad. Read it like the hiring manager who wrote it. List the 3 things this person most needs the hire to be good at, in order of importance. Then list the 3 tasks this role will actually spend most of its time on. Then list the exact skills and keywords that appear more than once. Do not write the letter yet.”
Then build your proof bank: 8 to 12 short bullets from your own work history, each in the form of tool plus scope plus result. “Cut onboarding time 30% by rewriting the welcome sequence for 400 new accounts” is proof. “A highly motivated team player with excellent communication skills” is not proof — it is the kind of line AI writes when you give it nothing to work with, and it is the first thing a recruiter’s eyes slide past. If you have not written a resume with AI yet, my guide to writing a resume with AI shows the same proof-first approach for the other document in the application.
Which AI Tool Writes the Best Cover Letter?
ChatGPT, Claude, and Gemini all handle this job well, and the differences between them are smaller than the difference between a good prompt and a lazy one. The table below reflects how each tool behaves with the prompt system in this guide, and the test underneath it is one you can run yourself in ten minutes.
| Tool | What it does best | Watch out for | Best for |
|---|---|---|---|
| ChatGPT | Clean structure and reliable length control; follows word-count limits better than most | Defaults to formal, corporate phrasing unless you tell it your tone | A first draft you can shape quickly |
| Claude | Natural, conversational tone; best at keeping your voice when you give it a writing sample | Longer drafts that need trimming to one page | Letters that need to sound like a person, not a template |
| Gemini | Strong at pulling keywords from the job ad and matching them to your proof bullets | Tends to pad with extra sentences about the company | Keyword matching for applicant tracking systems |
Run the same-notes test yourself: paste the same job ad and the same proof bank into all three tools with the draft prompt from the next section, then score each draft on three things. First, count the generic phrases — “I am thrilled to apply,” “passionate about excellence,” “dynamic team player” — and subtract a point for each. Second, check the word count: a draft over 350 words needs cutting, and a draft under 200 probably skipped your proof. Third, highlight every claim you cannot back up from your own history; each one is a sentence you will have to defend in an interview. The tool with the fewest generic phrases, the tightest length, and the fewest invented claims wins for that job. In my own writing sessions, Claude usually wins on tone and ChatGPT on structure, but the prompt matters more than the pick — a lazy prompt produces a lazy letter in all three.
One honest note on method: I did not run a live signed-in shootout across the three tools for this guide, and I am not presenting the table above as a lab benchmark. It reflects each tool’s documented behavior and the prompt system in this article. The test in the paragraph above is the real comparison — run it with your own job ad and your own proof, and trust your score, not mine.
How Do I Write a Cover Letter With AI, Step by Step?
Six steps, in order. The order matters: the facts come first, the AI second, and your judgment last.
- Decode the job ad (3 minutes). Run the prompt from the first section and keep the top 3 needs and the repeated keywords. Everything in the letter should trace back to this list.
- Build the proof bank (10 minutes, once). Write 8 to 12 bullets in tool-plus-scope-plus-result form. Reuse this bank for every application; only the job ad changes.
- Draft with constraints. Paste the job ad, the top needs, and your proof bank, then use this prompt:
- Run the specificity pass. Replace every sentence that would fit five other employers with a proof line and one line about this company: a product, a project, or a reason you want this team.
- Ask for a hiring manager’s critique. Prompt: “Do not rewrite the letter. Read it as a hiring manager for this role and tell me: which sentence is the weakest, which claim needs proof, and what would make you stop reading?” Fix what it flags; ignore the rest.
- Fact-check every line. Compare every date, title, tool, and number against your resume. AI invents details when it has gaps, and a single mismatch costs trust — and can surface in the interview, as my guide to preparing for a job interview with AI explains.
Here is the draft prompt for step 3. Copy it, fill in the brackets, and paste your proof bank underneath:
“Write a cover letter for the [job title] role at [company], using ONLY the facts below. If something is missing, ask me instead of making it up. Rules: maximum 300 words, 3 or 4 short paragraphs. Open with one specific line about this company, not a greeting. Each paragraph must connect one of my achievements to one requirement from the job ad. Use natural, confident language, not formal corporate phrasing. Do not use the words ‘passionate,’ ‘thrilled,’ or ‘dynamic.’ End with one sentence about what I would do in the first 90 days.”
What Does a Good AI Cover Letter Look Like?
Here is a worked example. The job is a Customer Success Manager role at a fictional mid-size software company, Northwind Analytics. The ad asks for someone to cut churn, run onboarding for new accounts, and work with the product team on feature requests. The applicant, Maya R., has three years in customer success at a smaller software firm. Everything in the letter below traces to those facts — nothing is invented, and the company detail in the opening comes from the ad itself.
“Dear Hiring Manager,
Northwind’s move from seat-based pricing to usage-based pricing is exactly the kind of change that tests a customer success team — and it is the work I have been doing for three years. I would like to bring that experience to your Customer Success Manager role.
At Brightpath Software, I owned onboarding for 400 new accounts a year. I rewrote the welcome sequence and cut time-to-first-value from 14 days to 9, which moved our 90-day retention from 71% to 82%. I also ran the weekly churn review, where I flagged at-risk accounts from product-usage data and recovered 11 of 14 accounts I escalated in my last year.
I know the other half of this job is the loop back to product. I logged and prioritized 60 feature requests a quarter, and two of the requests I championed — bulk export and SSO for small teams — shipped within two quarters. I am comfortable telling a customer ‘not yet’ when the answer is honest and the workaround is real.
In my first 90 days, I would rebuild the onboarding sequence around your new pricing model and set up the same usage-based churn review I ran at Brightpath. Thank you for considering my application — I would welcome the chance to discuss how I can help Northwind keep the customers its pricing change brings in.
Sincerely, Maya R.”
Notice what the letter does not contain: no “I am thrilled,” no “passionate team player,” no claim the applicant cannot prove. The opening line works only for Northwind. Every number — 400 accounts, 14 to 9 days, 71% to 82%, 11 of 14 — comes from the proof bank, not the AI’s imagination. And the letter is 268 words: short enough to be read in full, long enough to say something. If your draft does not look like this — specific company line, proof in every paragraph, nothing you cannot defend — run the specificity pass again.
Can Employers Tell If AI Wrote My Cover Letter?
Usually not by software — and that is not the risk you should worry about. Applicant tracking systems parse resumes and rank candidates against job requirements; no major ATS is documented to natively detect AI-written text. AI-writing detectors exist as separate add-ons, but they are unreliable in exactly the way that matters: OpenAI withdrew its own AI text classifier in July 2023 because of low accuracy — on an English test set it correctly identified only 26% of AI-written text — and a Stanford-led study found detectors flagged 61% of essays by non-native English writers as AI-generated while classifying US students’ essays as human almost without error.
People are the real detector. In a TopResume survey of 600 US hiring managers in June 2025, 19.6% said they would reject a candidate with a fully AI-generated resume or cover letter, while 52% said AI was acceptable for proofreading or drafting support. The same survey found only 33.5% of hiring managers could pick the AI-written resume out of four examples — barely above a 25% random guess. So most readers cannot prove you used AI, but about one in five will reject a letter that reads as fully machine-written. Greenhouse’s report adds the other side of the same coin: 65% of hiring managers have caught applicants using AI deceptively. The line is not “used AI” versus “did not use AI.” It is “used AI to draft, then made it true and personal” versus “sent the first draft untouched.” Never hide prompt injections or white-text keywords in a document — hiring managers report catching them, and being caught manipulating the process is a far worse outcome than being ranked lower.
What Should I Never Paste Into an AI Chatbot?
A cover letter needs less personal data than you think, and the chatbot needs even less. Before you paste anything, strip these out:
- Government ID numbers and full addresses. Your city and state are enough; the letter does not need your street address, and the chatbot never does.
- Salary history and current compensation. Nothing in a cover letter requires it, and it weakens your negotiating position if it leaks into a draft you reuse.
- Confidential details from your current employer. Client names, revenue figures, internal project names, and anything under an NDA stay out of the chat — paraphrase the achievement (“cut onboarding time 30%”) without naming who it was for.
- Your full resume as a file. Paste the proof bank instead. It gives the AI everything it needs and keeps the rest of your history out of the conversation.
- References’ names and contact details. Other people’s private information is never yours to paste into a chatbot.
Assume anything you paste may be stored by the provider, and delete the chat when the letter is done. If the tool offers a setting to exclude your chats from training, turn it on before you start — but do not treat it as a reason to paste sensitive data anyway.
How Do I Check the Letter Before I Send It?
Run this five-point check on every letter, no matter how good the draft looks. It takes five minutes and catches the errors that cost interviews:
- Every claim is on your resume. If the letter says you did it and the resume does not, one of them is wrong. Fix the document, not the story.
- One line works only for this company. Delete the company name. If the letter still reads fine, it is generic — add the product, project, or reason that ties it to this team.
- It fits on one page at 250 to 350 words. Shorter reads as thin; longer reads as unedited. Read it aloud: if you stumble, the sentence is doing too much.
- The keywords match the ad. The three top needs from your decode step should each appear in the letter, in the employer’s own words.
- Your LinkedIn profile agrees with it. A recruiter who likes the letter will open your profile next — keep the titles, dates, and numbers consistent. My guide to writing a LinkedIn profile with AI covers that half of the application, and if you want the whole workflow in one place, my guide to using AI for job applications ties the resume, cover letter, and tracking together.
Frequently Asked Questions
Yes, in most cases — as long as the letter is true and it is yours. University career centers at Harvard, Michigan State, and Carnegie Mellon all publish guidance on using AI for application materials, and their advice is consistent: let the AI decode the job ad, structure the letter, and polish the language, but bring your own achievements and rewrite the result in your own voice. One caveat from Harvard: read the application instructions first, because some organizations have their own rules about AI in applications. If an employer asks you not to use AI, follow their rule.
No. Applicant tracking systems parse documents and rank candidates against job requirements — no major ATS is documented to natively detect AI-written text. What an ATS does punish is a letter that misses the job’s keywords, so the keyword-matching step in this guide matters more than which tool wrote the draft. Keep the file simple: plain text or a standard PDF, no tables or graphics the parser can choke on.
Software cannot reliably prove it. OpenAI withdrew its own detector in 2023 for low accuracy, and Stanford research found detectors wrongly flag non-native English writing most of all. Recruiters, though, notice the signs of an unedited draft: generic praise, claims with no numbers, and sentences that would fit any company. In a TopResume test, only 33.5% of hiring managers picked the AI-written resume out of four — but 19.6% said they would reject a fully AI-generated application. Edit until the letter could only be yours.
Between 250 and 350 words — three or four short paragraphs on one page. That is long enough to connect two or three achievements to the job’s requirements and short enough that a hiring manager reads the whole thing. If your draft runs past 350 words, cut the weakest paragraph rather than shrinking the font. If it runs under 200, you probably skipped your proof bank and the letter is all adjectives.
You are not required to disclose it unless the employer asks. Most hiring managers accept AI for drafting and proofreading — 52% in TopResume’s 2025 survey — and object to letters that are fully machine-written and unedited. What you must never do is hide the help with tricks like prompt injections or white-text keywords: Greenhouse found 65% of hiring managers have caught applicants using AI deceptively, and being caught manipulating the process ends applications. If a job posting publishes an AI policy, follow it exactly.
Any of the three main tools works, and the prompt matters more than the pick. ChatGPT is the most reliable at holding a word-count limit, Claude usually produces the most natural tone, and Gemini is strongest at matching the job ad’s keywords to your achievements. Run the same job ad and proof bank through all three with the draft prompt in this guide, score them on generic phrases, length, and invented claims, and use whichever wins for that job.