Skip to content

Using AI to write your cover letter: what works and what gets you binned

By Helen Inman7 min read

The giveaway is almost never the grammar. AI writes perfectly serviceable sentences; that's the whole point of it. The giveaway is that three letters in the same pile open with the identical construction, praise the company's "commitment to innovation" and then say precisely nothing about what the candidate has ever actually done.

I've sat on the receiving end of stacks like that. You can read one in eleven seconds and learn nothing.

That doesn't mean you should refuse to touch AI. It means you need to know exactly which parts of the job it does well and which parts it quietly ruins.

Why the volume problem makes this urgent

The maths of the current market explains the behaviour on both sides. ONS figures for April to June 2026 put UK vacancies at 712,000, down 7,000 on the previous quarter, and payrolled employees fell by 85,000 in the year to May 2026. The House of Commons Library puts the squeeze plainly: there were 2.5 unemployed people for every vacancy in March to May 2026, the highest that ratio has been since the middle of the pandemic.

For younger applicants it's harsher still. Youth unemployment now stands at 16.4%, its highest level for 11 years.

So candidates apply for more roles, and generative tools make applying for more roles almost free. Employers then receive volumes they cannot read properly, and start hunting for reasons to reject rather than reasons to interview. Your cover letter has to survive that.

What UK hiring managers say about AI-written applications

Not much that's flattering. A survey of 625 UK-based hiring managers found 80 per cent viewed AI-generated CVs and cover letters negatively, and 74 per cent claimed they could spot AI-generated content.

Treat the 74 per cent with some scepticism. Plenty of people who insist they can spot it are actually spotting bland, which is a different crime and one humans have been committing unaided for decades. But the sentiment matters, because a hiring manager who suspects a letter is machine-written stops giving it the benefit of the doubt.

Meanwhile the employer side is far from fully automated. CIPD research found around a third of organisations use some form of AI or machine learning in recruitment and onboarding, and roughly 80% of SMEs and public sector organisations use no AI at all in hiring. If you're applying to a 40-person consultancy or an NHS trust, a human is reading your letter. Write for that human.

There's a live cost to over-automation on the employer side too: more than a third of recruiters say AI tools are now causing them to miss out on strong candidates. Nobody is winning this arms race. You just need to lose it less.

The four tells that mark a letter as machine-made

  • Adjectives doing the work of evidence. "Highly motivated", "proven track record", "passionate about excellence". None of these can be checked, so none of them count.
  • Praise the company can't verify you mean. If your admiration paragraph would work for any competitor with a similar website, it's filler.
  • Perfect symmetry. Three paragraphs of near-identical length, each with a tidy topic sentence. Real people write unevenly.
  • American English creeping in. "Resume", "organization", "I reached out to", "utilize". A UK recruiter clocks this instantly, and it signals you didn't reread the thing.

Use it as a researcher and an editor, not a ghostwriter

Here is the split that works. AI is genuinely good at analysing a job advert, spotting the competencies buried in corporate phrasing, tightening flabby sentences and telling you when a paragraph is repeating itself. It is bad at knowing what you did last March, and it will invent a plausible version if you let it.

So brief it like you'd brief a junior colleague. Paste the job advert, paste your CV, and ask it to list the three requirements the employer clearly cares most about and where your experience is thinnest against them. That's a genuinely useful five minutes. Then close the tab and write the middle of the letter yourself.

The middle is the bit that gets you interviewed: the paragraph where you name a project, a number, a problem you fixed. If you're unsure how the pieces fit together, our cover letter structure and template walks through what belongs in each section.

The prompt that produces something usable

Vague prompts produce vague letters. Compare "write me a cover letter for this marketing role" with something like this:

"Here is the job advert and my CV. I want a 250-word draft. Use only the facts in my CV. Use British English. No adjectives about my personality. Second paragraph must reference the two specific examples I've listed below. Plain, direct tone, no enthusiasm phrases."

You'll still need to rewrite roughly half of it. That's fine; you're now editing rather than staring at a blank page, and the structure of the argument is yours.

Feed it your evidence in raw form: "cut supplier onboarding from 14 days to 6", "trained four new starters", "ran the Leeds site through a CQC inspection". Facts are the one input AI cannot fabricate for you and the one thing a hiring manager will actually remember.

The ten-minute human pass before you send

Read the draft aloud. Anything you'd never say out loud in a meeting comes out.

Then do three things. Delete every sentence that could appear in someone else's letter unchanged. Add one detail only you could know, ideally something about the specific team, service or problem the role exists to solve. Check the spellings: British forms throughout, correct company name, correct hiring manager name if you have it.

If the letter still reads as generic afterwards, the problem isn't the software. Our guide on writing a cover letter that doesn't sound like everyone else's deals with that properly.

Should you tell them you used AI?

Only if they ask, and some now do. A growing number of employers include a line in the application about acceptable AI use, particularly in graduate schemes and the public sector. Read it and follow it, because being caught contradicting a declaration is a far bigger problem than the AI itself.

The honest position most employers hold: using it to draft and polish is fine, using it to claim skills you don't have is fraud. That line is not blurry.

One more practical point. Whatever you produce still has to get through screening software before a person sees it, and formatting matters as much as wording. If you haven't checked that side, read our piece on what applicant tracking software actually looks for.

The test I'd apply to every letter

Cover the top of your letter so your name is hidden, then ask whether anyone reading it could still work out it was you. If the answer is no, you've sent a form letter with better punctuation.

The fix takes ten minutes: one specific number, one named piece of work, one sentence explaining why this employer rather than the four others you applied to on Tuesday. AI can help you find those things faster. It cannot supply them, and the moment you expect it to, you're back in the pile of 200 that nobody finishes reading.

Sources