Step-by-step guide AI Job-Search Fast-Track: experienced professionals (3–15 years) →
Not reliably. No tool can prove that a CV, or resume, was written with ChatGPT or Claude. OpenAI withdrew its own detector in 2023 because it was inaccurate, and independent research has shown detectors wrongly flagging human writing. What readers do notice is generic wording, and some employers publish rules on what AI help they accept. The real risk is not detection. It is a claim the assistant made bigger than the truth, which you then have to defend in an interview. Here is what the evidence says, what named employers allow, and what we saw when we tested it.
Why no tool can prove it
Start with the company that makes ChatGPT. OpenAI released an AI text classifier on 31 January 2023. By its own figures, it correctly flagged 26% of AI-written text and wrongly labelled human-written text as AI 9% of the time, and it was very unreliable on texts shorter than 1,000 characters. Six months later OpenAI added a note to the announcement: as of 20 July 2023 the classifier was "no longer available due to its low rate of accuracy". A CV bullet runs to a line or two. That is far below the length at which even the maker trusted its tool.
Independent research found a second problem. Researchers at Stanford ran seven widely used detectors on 91 English essays written by non-native speakers for the TOEFL language test. On average, the detectors labelled 61% of those human-written essays as AI-generated, while they were close to perfect on essays by US school pupils (Liang and colleagues, revised July 2023). Those detectors are three years old and newer ones may do better. The lesson stands: writing in plain, careful English as a second language can look like AI to a machine. An employer who rejected candidates on a detector's verdict would be rejecting real people for how they write.
Screening software is built for a different job. The vendor documentation we read for our article on whether ATS reject CVs automatically describes matching a CV against one employer's criteria. None of those pages mentions scoring whether the text was written by AI.
What readers do notice
People notice sameness. In Greenhouse's survey published on 20 October 2026, 77% of 1,200 US job seekers had used AI in their search in the past year, and 60% had changed or considered changing how they write so they wouldn't be mistaken for AI. When most applicants use the same tools, the applications start to read alike.
Hiring managers are reacting to that, not to AI as such. In a Robert Half survey of more than 2,000 US hiring managers, run in November 2025, 65% said a surge in applications, many of them enhanced or generated by AI, had made candidates' skills harder to verify. Among the steps respondents had taken in response, 38% had increased the number of interviews per candidate and 32% had updated job descriptions to discourage generic AI answers (Robert Half, 10 March 2026).
Attitudes are split. In a TopResume survey of 600 US hiring managers, run by Pollfish in May 2025, about one in five (19.6%) said they would reject a candidate with an AI-generated CV or cover letter, 52% said AI is acceptable for proofreading or drafting support, and 14.5% said candidates shouldn't use AI at any stage (TopResume, 3 June 2025). You can't know which of those readers will open your application.
We saw what generic looks like in our own test. Given a fictional CV and an invented job ad, the plain request "write a cover letter for this job" produced about 330 words that opened with being excited to apply and gave a reason for wanting the job that could have been sent to any company. A prompt built on one real reason and one proof produced 118 words that only this candidate could have sent. You don't need a detector to tell them apart. One of them sounds like everyone else.
What employers' own rules say
A few employers publish rules for candidates. We read two on 10 November 2026: Anthropic's guidance for candidates, last updated on 10 July 2025, and the UK Civil Service's candidate guide to AI in recruitment.
| Employer | What it allows | What it doesn't |
|---|---|---|
| Anthropic, the maker of Claude | Write your first draft yourself, then use Claude to refine it and to polish how you describe your work. | AI-written answers about experience you haven't had. No AI in live interviews or take-home tasks unless the company says otherwise. |
| UK Civil Service | Refining and clarifying your ideas, research on the organisation, checking spelling, grammar and clarity, practising for interviews. | Inflating or inventing skills and experience, copying in generic responses, AI during live assessments. It may reject applications where AI is used inappropriately. |
The pattern is the same in both. Help with wording is accepted. Invented or inflated content is not, and live assessments are yours alone. The Civil Service also asks candidates to confirm that what they submit is true and accurate, which turns an inflated claim from a matter of style into a matter of honesty.
Other employers set different rules, and some set none. Read the careers page and the application form before you start. If a form asks whether you used AI, answer truthfully.
What can be detected: hidden text
One use of AI is detectable, and it isn't writing help. Some applicants hide instructions for the screening software in their CV, in white text, for example. A study published at the USENIX Security Symposium in 2026 examined about 200,000 real resumes collected by the company hireEZ, found hidden instructions in roughly 1% of them, and built detection methods that reached high precision in the authors' checks (Zhang and colleagues, May 2026). In a Greenhouse survey published on 19 November 2025, about 22% of US hiring managers said they had found hidden instructions in resumes (Greenhouse).
So the line employers can actually police runs between honest help and tricks. Stay on the right side of it. Anything a reader can't see on your CV shouldn't be in the file.
The real risk is a claim bigger than the truth
For our guide we created a fictional candidate, Robin Vermeer, and ran a full application routine in ChatGPT, with Claude as a cross-check. The prompts for the CV told the assistant never to invent anything, and it invented no numbers. It still drifted in every kind of document we asked for:
- The CV. While collecting Robin's achievements, the assistant made the wording bigger than the facts in four places. Tailoring the CV to one ad added three more overstatements, among them a phrase lifted from the ad that Robin's record didn't support.
- The cover letter. The generic letter included a claim about enjoying developing team leads that appeared nowhere in Robin's evidence.
- Interview answers. Asked to turn a result into a spoken story, the assistant kept every number and then invented motives for Robin's decisions, such as limited career progression. They sounded plausible. They were guesses.
None of this would be caught by a detector. All of it would be caught by a person. When we practised an interview with a deliberately vague answer, the assistant playing the interviewer asked at once for the numbers before and after, what had changed and what Robin's own part had been. A real interviewer asks the same three things. If the honest answer is smaller than the line on your CV, you have a problem that has nothing to do with which tool wrote it.
The fix, in short: collect your facts first, in one private list. Let the assistant rearrange and tidy only what is on that list. Then have a fresh chat, ideally in a second assistant, compare the finished CV with the list, line by line. Our article on tailoring a CV with ChatGPT or Claude has the truth-check prompt and what it caught in our test.
Five checks before you send
- Could you talk about every line for two minutes without notes? If not, the line is the assistant's, not yours.
- Is every verb the size of what you did? Look hardest at led, managed, owned, drove and identified. These are the ones that grow on their own.
- Does each claim carry a number or a proof you could show? If you think you have none, see how to quantify achievements when you have no numbers.
- Did the ad's phrases get in only where they're true? Assistants copy the employer's wording whether or not it fits you.
- Have you read the employer's rules on AI? Follow them, and answer truthfully if you're asked.
The whole routine, from the list of facts to the truth check, the cover note and interview practice, is in The AI Job-Search Fast-Track, experienced professionals edition, with eleven prompts tested in ChatGPT and Claude and a worked example you can follow line by line.
So use the assistant. It is good at making true things clear, and most employers accept that. Just make sure that everything it makes clear is true, because the person across the table will ask.
Frequently asked questions
Can employers detect if a CV was written with ChatGPT? Not reliably. OpenAI withdrew its own AI text classifier in July 2023 because of its low accuracy, and a Stanford study found that seven detectors wrongly flagged more than half of a set of essays by non-native English writers. What readers do notice is generic wording with no specifics, and claims that fall apart under an interviewer's follow-up questions.
Is it OK to use AI to write my CV? It depends on the employer, so check its rules. The published guidance we read from Anthropic and the UK Civil Service allows AI for refining and checking what you wrote, and forbids using it to invent or inflate experience or to answer live assessments. In a 2025 survey of 600 US hiring managers, 52% accepted AI for proofreading or drafting support and about one in five would reject an AI-generated CV or cover letter.
Do applicant tracking systems check whether a CV is AI-written? None of the vendor documentation we read describes that. The pages from Workday, Greenhouse and Ashby describe matching a CV against one employer's criteria for one job. Hidden text is a different matter: researchers have built detectors for instructions hidden in CVs, and hiring managers report finding them.
Should I say that I used AI if an application asks? Yes, answer truthfully. Some employers ask candidates to confirm that the information they give is true and accurate, and the UK Civil Service says it may reject applications where AI is used inappropriately. Using an assistant to tidy wording you can stand behind is very different from submitting claims you cannot defend.