When Saying You Used AI Helps, and When It's Noise

Should you mention AI use when nobody asked?

Only where it changes what the artifact proves. A resume claims you have certain skills, not that you personally typed every word describing them, so its evidence never depended on your drafting process. A portfolio piece makes a different claim: that the reasoning inside it is yours. Where that claim is live, saying what a model did and what you did changes what a reader can conclude. Where it isn't, saying so changes nothing but your own nerves.

Start from where most people actually are. In a 2026 Prospects survey of 4,990 UK students and early-career jobseekers, 73% said they had used AI somewhere in their applications ¹. Whatever you did probably sits inside that number already, which is a reason to stop treating unprompted disclosure as a confession and start treating it as a design choice: does saying this add information the reader can use, or does it just add words.

A middle case is worth naming too: a group project or a paper where AI was a background research tool rather than the finished product. There the honest move is usually a brief methodological note rather than a headline disclosure, the same way you'd footnote a source you consulted. Reserve the prominent, standalone disclosure for cases where the artifact's entire claim rests on unaided skill, and let the quiet cases stay quiet.

Keep quiet on a resume or cover letter

A cover letter that opens with a note about which parts a model helped write is answering a question the reader never asked, and it makes the letter about your process instead of about the role. In an experiment with US supervisors who had hiring experience, applicants who disclosed AI use in a letter of motivation were trusted less than applicants whose letters said nothing about how they were written ². It argues against volunteering a disclaimer nobody requested.

Nothing about that finding argues for lying if asked directly, which is a different situation entirely. Unprompted disclosure backfires here because a resume and a cover letter are already understood to be drafted, polished, and often reviewed by someone other than the candidate, model or human, so a note about AI assistance doesn't tell an employer anything they hadn't already assumed. What it does add is a flag that reads as anxiety, on a document where the actual claims, your job titles, your dates, the outcomes you describe, are what an employer is meant to evaluate. Keep the disclosure energy for the document where it's actually load-bearing.

Name the check on a portfolio, not the tool

A portfolio project is the opposite case, because the entire point of showing one is to demonstrate that you can produce this kind of work. Resume mentions of AI skills have grown more than ninefold since 2022, and two-thirds of those mentions now come from outside computer science majors ³, which means naming AI involvement in a project is no longer an unusual admission. It's becoming the normal way people describe how contemporary work gets made.

What earns credit isn't the disclosure itself, it's the specificity behind it. A line naming which parts a model generated, what you changed, and what you verified against something outside the conversation reads as evidence of judgment. A generic AI-assisted tag reads as nothing at all, because what a reviewer is actually trying to tell from a portfolio is which decisions were yours, and a vague label doesn't answer that. The same logic applies if you're ever asked to hand over the record of how you worked: what an employer is actually looking for in that record is specific acts, not a tool name.

The earlier resume advice and this advice can look like they contradict each other, and they don't. A resume names AI skills the way it names any other skill, as a general claim about you. A portfolio project demonstrates one specific instance of that skill in use, with a specific model, on a specific problem, and that specificity is exactly what makes the disclosure worth reading rather than worth skipping.

What if you followed a rule you never mentioned?

If an employer stated a rule and you followed it, you don't owe unprompted disclosure beyond what the rule itself required. Compliance with a stated instruction isn't the same act as choosing, unasked, to volunteer information about your process. A brief that says AI is welcome and asks for a short account of how you used it has already told you what to disclose. Following it fully answers the question, and nothing extra is owed.

Where things get genuinely uncomfortable is the reverse case, a rule you didn't follow. That isn't a disclosure decision at all anymore. Whether that crosses into dishonesty turns on the rule and what you did with it, not on how you phrase an unprompted mention now, and no amount of careful wording in a project's read-me fixes a broken instruction upstream of it.

Don't let disclosure become your compliance strategy

Being caught after the fact costs more than saying something up front. In the one experiment testing exactly this pattern, set between a professional and his client rather than in hiring, the professional whose AI use was exposed by someone else was trusted less than one who had disclosed it himself, who was trusted less than one who said nothing at all ⁴. What transfers to hiring is that ordering, not the exact size of the gaps.

Disclosure isn't free, but it isn't the worst outcome on the table. Confusing those two costs, the cost of telling and the cost of being caught hiding, is what makes this question feel harder than it actually is. The honest shortcut is to stop asking whether you should disclose as one general question and start asking what the document in front of you is actually claiming. How employers themselves choose between asking, checking or watching runs on the identical logic from their side of the process: match the method to what the stage is meant to measure, and the right amount of disclosure for a candidate tends to fall out of the same reasoning.

Read the evidence >