Usually not reliably — and that is not the good news it sounds like. The real risk is being wrongly accused, and it falls hardest on people who did nothing wrong.
Nearly half of jobseekers now use AI to build or improve their CV — 45% in a Canva/Sago survey of 5,000 people across eight countries. So this is a mainstream question, and most pages answering it are published by companies selling CV tools, which gives them an obvious reason to reassure you.
There is no reliable way to prove a CV was written with AI. Detection tools infer it from statistical patterns — how predictable the word choices are — and they are wrong often enough that a serious employer cannot safely act on them alone.
Independent testing puts typical detector accuracy somewhere between 65% and 90% depending on the tool, with false-positive rates on some platforms measured between 15% and 45%. Newer detectors are markedly better — 2025 research from the University of Chicago Booth School found one tool with a near-zero false-positive rate — but "some tools are now good" is not the same as "employers reliably detect AI", and CVs are short, formulaic documents, which is the hardest case of all.
These tools do not fail randomly. They flag writing that is predictable — simpler vocabulary, shorter sentences, conventional structure. That describes a great deal of perfectly honest writing, and in particular it describes writing by people whose first language is not English.
Research has repeatedly found detectors misclassifying non-native English writing as AI-generated, with false-positive rates in some studies exceeding 20% for that group. Formal, formulaic passages are flagged disproportionately too — and a CV is nothing but formal, formulaic passages.
Read that in the context of an Indian graduate applying to UK employers. The tool is not measuring dishonesty. It is measuring how closely your prose resembles a statistical average, and penalising you for writing clear, plain, correct English as a second language. If a detector is used as evidence, the people most likely to be wrongly accused are the ones with the least standing to argue.
The serious ones have largely given up on detecting AI in documents, because the question turned out to be unanswerable and the wrong one anyway. Whether you used a tool to phrase a bullet point does not tell them whether you can do the job.
What they do instead is stop trusting the document. That is the whole shift toward assessments, work-sample tasks and structured interviews — and it is why "can they tell?" is fast becoming irrelevant. In the same period, 22% of jobseekers began using bots to auto-apply, and 28% admit generating fake work samples. Employers reacted to the volume, not to the prose style.
It cuts the other way too. Indian background-check data shows 17% of candidates had a résumé mismatch and 13% failed employment verification. Verification is spreading because claims of every kind became cheap, not because of AI writing specifically.
Using AI to write your CV is not misconduct. Using it to claim things that are not true is, and that is what background checks and assessments catch — reliably, unlike detectors.
Keep the facts yours. Let a tool tighten your phrasing, then check every date, title, employer and number yourself. The most common way AI-assisted CVs cause real trouble is quiet inflation: "led" instead of "worked on", a scope enlarged, a date rounded.
Keep your own voice in places that matter. Not to defeat detection, but because a covering note that sounds like a person is more persuasive than one that sounds like everyone else's.
Carry something checkable. This is the only durable answer. A verified assessment score, a portfolio, a reference who will pick up the phone — evidence that does not depend on anyone believing your prose. If you are ever wrongly flagged, being able to point at a proven result ends the conversation faster than any argument about writing style.
Ask what the claim is based on. If the answer is a detector score, it is fair to say the tools are known to be unreliable on short formal documents and known to misclassify non-native English writing — and to offer something concrete instead: a task, a test, a reference. An employer that will not accept demonstrable evidence over a probability score has told you something useful about itself.
A Skilyo Verified badge comes from a timed, integrity-monitored assessment. It is evidence no writing-style tool can argue with.
Related: Why skills-based hiring is replacing resumes — the shift this is part of. And why you are not hearing back.
This is general career guidance, not legal or HR advice. AI-CV usage is from the Canva/Sago 2024 survey of 5,000 job seekers across eight countries. Detector accuracy and false-positive ranges are drawn from independent testing and peer-reviewed work on AI-writing detection, including "AI writing detectors are ineffective, unreliable and harmful" (English Teaching: Practice & Critique) and 2025 University of Chicago Booth School testing, which found detector performance varies widely and that at least one tool now achieves a near-zero false-positive rate. Auto-apply and fake work-sample figures are from the Greenhouse 2025 Workforce & Hiring Report (2,200 jobseekers). Background-check discrepancy rates are from AuthBridge's aggregated candidate checks, October 2024 – March 2025; AuthBridge does not disclose a sample size. Current as of August 2026.