If You Barely Use AI, Say So and Bring the Judgment

Do Not Fake Fluency

Faking fluency fails at the second question. An interviewer who asks how you use AI will usually ask what that looked like on a real task, or what you would check before trusting the output. A rehearsed line about being "AI-forward" has nowhere to go once that lands, and the gap between the claim and the answer does more damage than the honest version ever would.

Say plainly and calmly where you actually stand right now. "I haven't used it much day to day" is a complete sentence, not a confession, and it sets up the part of the answer that actually matters.

Self-rated skill and demonstrated skill are not the same measurement. In a study of 288 Taiwanese teachers who took both a self-report and a knowledge-based AI literacy test built on the same framework, correlations between the two ran from 0.07 to 0.24, and the profiles ran in both directions: some teachers overrated their own skill and some underrated it. ¹ That was a classroom sample rather than a hiring one, and the point that carries is narrower: a self-rating and a real example are different kinds of claim, and only one of them can be walked through when someone asks for specifics.

Describe the Judgment You Already Have

The transferable claim is not tool experience. It is judgment about when a tool, a person, or a process is trustworthy, and most experienced candidates have already exercised that judgment on something other than a model: a vendor's numbers, a junior colleague's draft, a report from a team you do not fully trust yet.

Say where you have used AI, even lightly, and where you have deliberately chosen not to. Name what you would want to verify before trusting a model's output in your specific field, and why that particular thing is the risk: a stale regulation, a number that needs to tie back to a ledger, a claim about a person that needs a source. That answer demonstrates exactly the behavior what good AI use looks like describes employers scoring for: framing the problem, demanding a source, and testing something against the world outside the conversation. None of that requires that you have used the tool constantly.

Make the parallel explicit if the interviewer does not draw it for you. If you have caught a factual error in a vendor deck, pushed back on a number a colleague handed you, or double-checked a figure before it went into something that mattered, say so and connect it forward: that is the same check, aimed at a different source. An interviewer who hears the mechanism described clearly does not need you to have already run it against a model.

Why a Fast Certificate Doesn't Fix This

A weekend certificate is weak evidence, and reading what one actually asks shows why. Google Cloud's Generative AI Leader exam, aimed at anyone in any job role with or without hands-on technical experience, runs ninety minutes, fifty to sixty multiple-choice questions, no prerequisites, for a fee. ² Two of its four content areas are that vendor's own product line. Passing it evidences familiarity with vocabulary, not judgment in the interviewer's field.

What carries more weight is what you actually did in the weeks before the interview, not a document saying you passed something. In a discrete choice experiment with 543 US hiring managers and HR specialists in December 2025, moving a hypothetical candidate's relevant work experience from zero to two years raised their probability of being selected by 21.4 percentage points, and the authors conclude that experience and skills weigh substantially more than how a degree was delivered. ³ Those were stated choices between profiles rather than real hires, and the study measured degrees rather than AI certificates, so read it as a direction and not as a rate: what you can demonstrate outweighs the packaging around it.

Name the Real Ceiling Honestly

Some roles genuinely need daily, hands-on AI use from day one, and no interview framing changes that. If the posting is explicit about that bar and your experience is thin, say so plainly rather than talking around it, and ask what ramp-up looks like on their side. An honest mismatch named early costs less than a mismatch discovered after an offer.

Most roles are not that. A role that only occasionally touches AI-assisted work is testing whether you would pick it up sensibly, not whether you already have. Is it a dealbreaker if your best candidate doesn't use AI? is the account from the other side of that same table, and it draws a narrower line than most candidates assume: refusal and unfamiliarity are read as two different things, and unfamiliarity closes in weeks.

What to Say If You Get the Follow-Up

If pressed on why your use is light, a true, specific reason beats a defensive one. "My last role didn't call for it much, and I was cautious about the parts I couldn't verify" describes judgment. "I haven't gotten around to it" describes drift, and it invites the interviewer to wonder what else you have not gotten around to.

If you have already started closing the gap, name what you actually did rather than a course title: a real task you used a model on, what you checked, what changed. One genuine week of practice on work you can describe in specifics gives you more to say than a certificate you cannot walk through, because it leaves you with a decision to describe rather than a completion to report.

Employers say they want people who can tell when AI is wrong. How do I show that? covers how to build and describe that evidence, even from a small amount of real practice, which is worth more here than another certificate.

See how it works >