How to humanize an AI-written cover letter
Add one specific thing only you could say. AI cover letters fail because they are interchangeable — swap the company name and the letter still works. Name the actual product, the actual problem, or the actual reason you applied. One concrete sentence outperforms three paragraphs of enthusiasm.
Typical length · 250–400 words. Comfortably inside the free plan. · Last reviewed 16 August 2026
Before and after
Before
I am writing to express my strong interest in this position. I believe my skills and experience make me an excellent candidate for this role. I am passionate about contributing to your team's success.
After
I am applying because your onboarding flow is the only one I have used that explains why it needs each permission. I spent two years rebuilding ours after a launch that did the opposite, and I would like to work somewhere that already gets this.
Generic enthusiasm became a specific observation about the employer plus relevant experience. The rewrite is longer and it earns the length — it could not have been sent to any other company.
The interchangeability test
There is one question that decides whether a cover letter is working: could this letter be sent to a different company by changing the name at the top? If yes, it is not doing anything, however well written it is.
That is the specific failure of a model-written letter. It has no knowledge of the employer beyond what the job advert contains, so it produces enthusiasm in general — strong interest, excellent candidate, passionate about contributing. Each sentence is fine and the letter as a whole is addressed to nobody.
The rewrite above passes the test in one clause: it names the company's onboarding flow and what is unusual about it. That sentence cannot be reused anywhere else, which is exactly why it is worth the space it takes.
One specific thing beats three general ones
The instinct when a letter feels thin is to add more claims. It rarely helps, because the reader is not counting claims — they are looking for evidence that you know what the job is and can do it.
One concrete observation about the employer, connected to one thing you have actually done, outperforms a paragraph of adjectives. In the example that connection is explicit: their onboarding explains its permissions, and the writer spent two years rebuilding one that did the opposite. The relevance is demonstrated rather than asserted.
It also gets you past the register problem. A letter built on a specific observation sounds like you because the observation is yours; a letter built on adjectives sounds like everyone.
Register is industry-specific, and rewriting can break it
The same letter is wrong for a law firm and a seed-stage startup, and this is where an automated rewrite most often overcorrects. Asking for something less formal can produce a letter that reads as casual in a context where formality is the signal that you understand the work.
Read the company's own writing before you decide the register — their careers page, their documentation, the job advert itself. Match roughly what you find there rather than picking a tone in the abstract.
One page remains the limit whatever the tone. A rewrite usually lengthens text, so check the count afterwards and cut the sentences that restate rather than the ones that specify.
Applicant tracking systems are not AI detectors
These get conflated constantly. Most applicant tracking systems parse and index applications so a recruiter can search them; some score for keyword match against a requisition. That is a different job from estimating whether text is machine-generated, and a keyword score says nothing about authorship.
Some employers have begun asking candidates to declare AI assistance, which is a policy question rather than a detection one. If a form asks, answer it — an undeclared rewrite is a much smaller problem than an inaccurate declaration.
The thing that actually filters at this stage is a person reading quickly, which brings it back to the interchangeability test.
Formats with the same problem
CVs and résumés — the adjective-to-number edit, applied to bullets instead of paragraphs.
Personal statements — the same voice problem when the subject is you and the stakes are a place.
Questions
- Can employers tell if a cover letter was written by AI?
- A recruiter reading quickly often can, and not because of a tool. Interchangeable enthusiasm, uniform sentence length and no specific knowledge of the employer are recognisable at speed. That is the same signature a detector responds to, noticed by a person with a stack of applications.
- Do applicant tracking systems detect AI writing?
- Most parse and index applications so recruiters can search them, and some score keyword match against the job requisition. That is not AI detection, and a keyword score says nothing about who wrote the text. Treat the two as unrelated systems.
- Should I declare that I used AI on an application?
- If the employer asks, yes. Some now include the question on application forms, and an inaccurate declaration is a far larger problem than an undeclared rewrite. Where nothing is asked, the judgement is yours — but the specific detail in the letter still has to be true.
- How long should a cover letter be after rewriting?
- Still one page, typically 250-400 words. Rewriting usually lengthens text, so check the count afterwards and cut sentences that restate what you have already said rather than the ones carrying the specifics — those are what the letter is for.
What to watch for
- The specific detail has to be true. A rewrite cannot invent your experience.
- Match the register of the industry — the same letter is wrong for a law firm and a startup.
- Keep it under one page regardless of what the rewrite produces.
If your writing gets flagged
Rewriting for rhythm and specificity tends to lower detection scores, because that is what detectors read as human. It is not a guarantee — detectors disagree with each other and change without notice, and we do not promise a result from any of them.