Humanetext

Why Your Cover Letter Sounds AI-Generated

Cover letter conventions produce the exact patterns recruiters now read as machine-written. Which habits to drop, and what to write instead.

Practical guidesBy Humanetext Editorial5 min read

The short answer

Cover letters read as machine-written because the conventional format teaches the same habits models overproduce: a formulaic opening, enthusiasm asserted rather than shown, competencies listed without evidence, and phrasing recycled from every template online. The fix is specificity. Name the actual role and one concrete thing about the company, replace adjectives about yourself with a result and a number, and cut every sentence that would fit any other application unchanged.

You wrote it yourself. You spent an hour on it. And reading it back, it sounds exactly like the thing a model produces when you ask for a cover letter.

This is not a coincidence, and it is not a failure of effort. The cover letter is one of the most heavily conventionalised documents in existence, and language models learned those conventions from hundreds of thousands of examples. When you follow the format you were taught, you produce the same output the model does, because you are both drawing on the same template.

Which means the fix has almost nothing to do with AI. It is the fix that has always made applications better.

Four habits, all taught as good practice

The formulaic opening. I am writing to express my interest in the Marketing Coordinator position advertised on your website. The reader knows what this is. They opened an application. The sentence spends your best line restating the envelope.

Enthusiasm asserted rather than shown. I am incredibly excited about the opportunity to join your dynamic team. Anyone can type this, which is precisely why it carries no information. Enthusiasm is demonstrated by knowing something specific about the organisation, not by claiming to feel it.

Competencies listed without evidence. I am a highly motivated self-starter with excellent communication skills and strong attention to detail. This describes no one. It is a list of adjectives that every applicant claims and none can be denied.

Phrasing recycled from every template. Leader in the industry. Fast-paced environment. I believe my skills would be a valuable asset. These phrases exist in every guide, every sample, and consequently in every model's training data.

Each of these was taught as correct. Each is also what a model overproduces, for the same reason: they are the safest, most common way to fill the space.

What to cut

The opening sentence. Almost always a restatement of why you are writing. Delete it and start with something only you could say.

Anything that restates your CV. The reader has it attached. A letter that lists your last three jobs in prose form has spent a page on information they already had.

Adjectives about yourself. Motivated, passionate, detail-oriented, hardworking. Replace each with the thing that would let a reader conclude it.

"Leader in the industry" and its relatives. Compliments about the company that could be pasted into any other application read as filler, because they are.

The closing formula. I would welcome the opportunity to discuss my application further. Of course you would. You applied.

What to write instead

The useful shape is three or four short paragraphs doing three specific jobs: why this role, what you have actually done, and what you would do here.

Before: I am writing to express my interest in the Content Manager position at your company. I am a highly motivated marketing professional with excellent written communication skills and a passion for storytelling. I believe my experience would make me a valuable asset to your dynamic team. I would welcome the opportunity to discuss my application further.

After: I have been reading your customer research posts since the pricing series in March — the one arguing that most churn surveys ask the wrong question stayed with me. I spent the last two years running content at a B2B software company, where I rebuilt the blog around interviews with actual users and took organic signups from 40 to 190 a month. I would want to do something similar here: less volume, more reporting.

Seventy-one words to eighty-three, so barely shorter. But the second one demonstrates the enthusiasm the first one claimed, gives a number, and states a view about how the work should be done. It also could not have been sent to anyone else, which is the whole test.

The general principles are the same ones that improve any prose: be specific, cut hedging, use verbs rather than abstract nouns. They are covered at length in how to write like a human and, for the abstraction problem specifically, in concrete writing.

The one test worth applying

Read your letter and ask: could I send this to a different employer by changing the company name?

If yes, it is not finished. Not because it is dishonest, but because it demonstrates nothing. The entire value of a cover letter over a CV is that it can show you understand this particular job at this particular place. A letter that survives a find-and-replace has thrown that away.

This is also, incidentally, the most reliable thing a recruiter actually notices. Detectors are unreliable and inconsistently used. Genericness is obvious to anyone who reads fifty applications a week.

If a detector is involved

Some employers now run applications through AI detectors. It is worth knowing what that does and does not mean.

These tools estimate how statistically predictable text is. They do not identify authorship and cannot produce a source. Clean, conventional, well-structured writing sits close to the statistical centre and scores as machine-like for that reason — which is exactly what a cover letter written to the standard format is. The mechanism is set out in how AI content detectors actually work.

So the same advice serves both purposes. A specific letter, full of things only you could write, is both more persuasive to a human and less like the statistical average. That is not a trick to fool a detector. It is what the detector was a bad proxy for.

If a form asks directly whether you used AI, answer honestly. An inaccurate answer discovered later is far more damaging than the use itself, and this is the same reasoning as when and how to disclose AI use.

Formal does not mean empty

The common objection is that this reads as too casual for a job application. That confuses two things.

Formality is how much distance the situation calls for. A letter to a law firm needs more than one to a four-person startup, and judging that is ordinary social skill.

Register is whether a person is discernibly behind the writing. That is available at every level of formality. A letter can be entirely correct and still sound like someone.

When a recruiter says an application felt generic, they almost never mean it was too formal. They mean nobody seemed to be there. The same distinction applies to work emails, which fail in exactly this way for exactly this reason.

Why this matters beyond style

A cover letter that names nothing and evidences nothing does not just read badly. It fails at its job.

The reader cannot tell why you want this role rather than any role, whether you understand what the work involves, or what you would actually do in it. So they skim it, move to the CV, and the letter has cost you a page of attention and returned nothing.

Specificity is not a stylistic preference here. It is the difference between a letter that gets you a conversation and one that gets skipped.

Common questions

Can employers detect AI in cover letters?
Some run applications through detectors, and those detectors are no more reliable here than anywhere else - they estimate statistical predictability, and a polished conventional cover letter scores high whoever wrote it. What recruiters detect far more reliably is genericness: a letter that could have been sent to any employer reads as unserious regardless of how it was produced.
Is it bad to use AI to write a cover letter?
Most employers have no stated policy and would not object to help with phrasing. The practical risk is different from the ethical one: an AI-drafted letter is generic by default, and generic letters get skimmed. Whatever tool you use, the specifics that make a letter land - the role, the company, your actual results - are the part you have to supply.
Why does my cover letter sound generic?
Almost always because it contains claims rather than evidence. I am a highly motivated team player with excellent communication skills describes nobody in particular. The same letter becomes specific the moment you replace each adjective with something that happened: what you did, where, and what changed as a result.
What should I not put in a cover letter?
Cut the opening formula about writing to express interest, any sentence restating your CV, adjectives about your own character, references to the company being a leader in its field, and the closing about welcoming the opportunity to discuss further. All are ritual, all appear in every template, and none tell the reader anything.
How long should a cover letter be?
Three to four short paragraphs, comfortably under a page. Roughly 250 to 350 words is plenty for most applications. Length is not the point though - a tight 200-word letter with two concrete results beats a full page of competencies, because the reader is scanning for evidence you can do the job.
Should I mention that I used AI to help write my application?
Only if the employer asks, and some now do. Where a form includes a question about AI use, answer it honestly - an inaccurate answer is far more damaging than the use itself. Where nothing is asked, help with phrasing is ordinary editing and no norm currently treats it as disclosable.

Keep reading