Can You Tell If a CV Was Written by AI

Cover letters and CVs polished with ChatGPT now hit every recruiter's inbox. The question isn't whether AI is being used — it almost always is — but can you tell if a cv was written by ai at the point of shortlisting. This guide is for HR managers and hiring leads who want a workflow that filters honestly without screening out strong candidates.

What detectors can and can't catch#

The short answer to can you tell if a cv was written by ai is yes, with caveats — and the caveats matter more than the headline.

ChatGPT output, especially from default GPT-4 settings, has a recognisable rhythm: balanced sentence lengths, frequent triplet structures, predictable transitions ("Moreover", "Furthermore", "In conclusion"), and a slight preference for hedging language. Most modern detectors catch raw ChatGPT output with high accuracy. The risk to students isn't that ChatGPT is undetectable — it's that any text edited *near* a ChatGPT draft starts to inherit those rhythms.

Detectors miss text when one of these things is true: the AI output has been substantially rewritten by a human; the text is very short (under about 80 words); the text is in a heavily technical domain the detector's training data didn't cover; or the writer used a humaniser tool plus their own edits.

Detectors over-flag text when the writing is unusually clean, formal or evenly-paced — which is why polished academic prose, ESL writers and well-edited essays get caught in false positives more often than they should.

No detector is right 100% of the time, and any vendor who claims otherwise is selling you something. The honest framing is that detectors are *indicators* — they tell you something looks unusual. The decision about what to do next is always human. Treat any score as a question, not an answer.

A practical screening workflow#

For HR and recruiters, the practical question isn't really "is this CV AI-written?" — half of cover letters now pass through ChatGPT for polishing. The question is whether the *substance* (specific projects, named technologies, dates, results) holds up against the cover letter's claims. Use a detector to flag candidates whose application is suspiciously uniform, then test the substance in the interview.

Practically: run the cover letter through a detector, but treat anything between 30% and 70% as "indeterminate, ask interview questions about the specifics in the letter". Anything above 70% combined with vague specifics is worth probing. Anything below 30% is fine — don't let detector noise screen out genuine candidates.

Common questions about can you tell if a cv was written by ai#

Should we reject candidates whose cover letters score high for AI?

No. A high AI score on a cover letter just means they used ChatGPT to polish it — which most candidates now do. Use the detector to identify candidates whose application is suspiciously generic, then test substance in the interview.

How accurate are detectors on cover letters specifically?

Less accurate than on longer essays, because cover letters are short, formal, and use predictable phrases by convention. Treat scores on text under 200 words as indicative only.

Is it legal to screen candidates using an AI detector?

In the UK and EU, yes, provided the detector isn't the sole basis for rejection and you handle the candidate's text under your privacy policy. As a matter of best practice, disclose in your hiring process that you may screen applications for AI use.

Try Exolio AI today#

You don't need to take our word for it. Run any piece of writing through Exolio AI and you'll get a clear AI vs human score, sentence-by-sentence highlighting on paid plans, and the ability to correct false positives so the model gets better for everyone.

**Try Exolio AI free — check your essay in seconds at app.exolio.com.**

If you've been falsely flagged by another detector, this is the fastest way to get a second opinion you can actually defend.