Do Recruiters Reject AI-Written Resumes?

This question gets asked in every job-search forum, and it always gets the same two useless answers: "recruiters can always tell" and "nobody cares, use it for everything". Both are wrong, and the truth in between is genuinely worth understanding — because it tells you exactly how to use these tools without getting burned.

Can a recruiter actually tell?

Reliably? No. AI-detection tools are not accurate enough to act on — they routinely flag ordinary formal, non-native, or simply well-edited writing as machine-generated. No serious hiring team is running your resume through a detector and binning you on the output, because doing so would mean rejecting a large number of real candidates on a coin flip.

But here's the thing. Recruiters don't need to detect AI, because AI writing announces itself in a much cruder way: it all sounds the same. A recruiter reading their 60th application of the day is not thinking "was this written by a model?". They're thinking "I have read this exact sentence before."

What is the actual tell?

Unedited AI output has a distinctive failure mode: it is fluent and empty. It produces grammatically immaculate sentences that assert competence without evidencing it. Compare:

Spearheaded cross-functional initiatives to drive operational excellence,
leveraging data-driven insights to optimise key business outcomes and
foster a culture of continuous improvement.

Cut checkout drop-off from 34% to 21% in one quarter by rebuilding the
payment step; shipped with two engineers and a designer, and the change
now handles ~40k transactions a week.

Both were written by an AI. The second one is good — not because a better model wrote it, but because a human brought the numbers. The first is what happens when you ask a model to write your resume for you instead of with you.

The problem was never that a machine wrote your resume. It's that nobody told the machine anything worth writing about.

What's the risk that actually matters?

Not detection. Inflation. Ask a language model to make your experience sound more impressive and it will happily oblige, because that is literally the instruction. You wrote "helped with the migration"; it returns "led the end-to-end migration of a distributed data platform serving 12 million users." You didn't lie — but the document now does.

Then you get the interview, and someone who has actually led a migration asks you what your rollback strategy was. This is where AI resumes fail: not at the parser, not at the recruiter's desk, but forty minutes into a conversation you can't sustain.

How should you actually use AI on a resume?

Think of it as an editor, not a ghostwriter. Editors are extremely good at the things people are bad at — being concise, being consistent, cutting a clause, finding the right verb, matching the register of a job description. Editors are not good at knowing what you did. That part is yours.

1. 1Write the ugly version first. Dump the real facts in plain language, with the actual numbers, however clumsily. made the reports faster, like 10 mins to under 1.
2. 2Hand it the job description. Ask it to align the vocabulary of your real experience to the role's language — without adding anything.
3. 3Give it an explicit prohibition. "Do not add any metric, tool, scope or outcome I did not state." Models follow this instruction well when you give it.
4. 4Read every line back and ask: is this still true? Downgrade anything that drifted. This is the step people skip, and it is the only one that matters.

Used this way, AI does the thing it's genuinely superb at — turning honest, badly-phrased truth into clear, well-phrased truth — and none of the thing it's dangerous at. Your resume ends up sounding like the best version of you, which is the only version that survives an interview.