Can employers tell if you used AI on your CV or cover letter?

Can they actually detect it?

Sometimes, partly, and not with proof. That is what the evidence supports.

OpenAI, the company behind ChatGPT, built its own AI text detector and then shut it down. Its notice reads: "As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy." By OpenAI’s own published numbers, the tool correctly identified just 26% of AI-written text, and wrongly labelled human writing as AI 9% of the time.

The false positives are not a fringe problem. In OpenAI’s own guidance for educators, it says AI detectors "sometimes suggest that human-written content was generated by AI," and notes its former detector flagged Shakespeare and the US Declaration of Independence as AI-generated. A 2023 Stanford study ran 91 essays written by real people, all non-native English speakers, through seven widely used AI detectors: on average the detectors wrongly flagged 61% of them as AI, and 89 of the 91 were flagged by at least one detector, while the same tools scored near-perfectly on essays by native-speaking US students. The pattern behind that result matters here: detectors tend to flag plain, predictable, low-variety wording.

To be fair to the other side of it: a 2025 review by Jisc’s National Centre for AI found that mainstream paid detectors now perform reasonably well on text that is entirely AI-generated and unedited. But the same review notes they are relatively easy to circumvent, they still produce false positives even at their best (around 1 to 2 percent, which at any real volume of applications is a lot of wrongly accused people), and detection on its own cannot prove AI use.

One detail matters for job seekers specifically. OpenAI warned that detection tools disproportionately misflag writing that is "particularly formulaic or concise," and writing by people using English as a second language. A cover letter is short and formulaic by nature, which is exactly the kind of writing these tools tend to trip on.

So the picture is this: a human-written letter can get flagged, an AI-drafted one can pass, and detection cannot prove either way from the text alone. Chasing "undetectable" is the wrong game entirely, for the reason below.

Detection is not the risk. This is.

The danger was never that a tool flags your writing. It is that you put something in your application you cannot stand behind.

An interview is where invented claims go to die. If your cover letter says you "led a team of ten" and you led three, no detector is needed to catch you: the follow-up question does. "Tell me about leading that team" is a wall you walk into at the worst possible moment, in the room, with the job on the line.

That is the real risk of using AI badly. Not being caught using it. Being caught out by it, when the letter made a promise your experience cannot keep.

The honest way to use AI that actually helps you

Used well, AI is not a way to manufacture experience you do not have. It is a way to say the experience you do have more clearly.

• Ground each claim in something true. If it is not on your CV or in your real work, it does not go in the letter.
• Tailor to the job using what you actually bring, and name genuine gaps plainly instead of papering over them. Naming a gap honestly reads as self-awareness; an inflated claim reads as a liability the moment it is tested.
• Ask one question of each line before you send it: if they ask me about this in the interview, can I back it up? Use AI freely for the lines that pass, and not for the ones that do not.

What good AI help looks like

This is exactly why we built Rolebird the way we did, and we will be upfront that this is our tool: it writes only from what your CV and the job ad actually support, and it strips out claims it cannot ground in your own documents. It also shows you the suggestions it refused to make because it could not back them from your evidence.

The principle stands whatever you use, though. Free tool, paid tool, or none: ground it in the truth and you will not have to worry about the question in this article’s title.

The question to stop asking

Stop asking "will they know I used AI." Start asking "can I defend each line if they ask." Use AI for the first job all you like, and do not let it do the second.

Common questions