Policy
The Application Asks If You Used AI: Answer It Straight
When a job application asks whether you used AI, tell the truth. The box is low stakes to fill in and high stakes to get wrong, since most employers treat application accuracy as part of the offer. If there's a free-text field, name the check rather than the tool: what you drafted, what you verified, what's genuinely yours. If it's a plain yes or no, answer honestly and bring the detail to the interview. A false attestation causes a cleaner problem than any true answer about AI ever would.
The takeThe tempting move is the quiet no, on the theory that nobody can check. That folk wisdom mistakes an unenforceable rule for a safe one. Detectors misfire too often for any employer to verify the box at submission, which is exactly why it works as an attestation instead: something you're on record stating. A false statement doesn't need to be caught to matter. It just needs to be relied on, and a signed application usually is.
Where Olive fits
Open a role and see what the work shows
If the invite comes from Olive, there's no box to guess at: the assignment states that an AI assistant is expected, and you describe what you did with it in a short written debrief, in your own words rather than a checkbox. You're granted the identical human-written report your employer receives, on every tier, at no cost.
Rank your shortlistWhat should you put in the AI-use box on an application?
Put the true answer, and treat the box as part of the application rather than as a trap to outsmart. Most employers asking this question are checking accuracy more than screening for AI use. Honesty is less exposed than it feels: in a 2025 Institute of Student Employers survey of UK early-careers employers, half (49%) reported no problem with candidates using AI in recruitment 1. Answering falsely puts something untrue on a document an offer is likely to reference later.
You are also not unusual for having something to disclose. In a 2026 Greenhouse survey of 1,700 jobseekers across the UK, Ireland and Germany, 78% said they use AI to tailor a CV or application at least some of the time 2. A box built around that reality is collecting a normal fact about how people write applications now, and answering it accurately is closer to filling in your location than confessing to something.
The stakes sit downstream, not in the box itself. A false attestation is a problem of record: if the form makes you confirm the information you gave is true, and that confirmation wasn't, you have created a record a company can act on months later, once you're already in the seat. What happens when an answer given now surfaces after an offer is out is a harder version of this same question, and it gets easier to answer when the record from this earlier stage was accurate to begin with.
Write the free-text answer around the check, not the tool
When a form gives you room to explain, name what you verified rather than which product you opened. A sentence describing what you drafted with a model, what you checked against a source, and what you rewrote in your own words tells an employer something they can act on. Naming the tool by brand tells them almost nothing they didn't already assume, and it reads as either an apology or a flex depending on the reader's mood that day.
Two published examples show what a good answer sounds like. Anthropic's own candidate guidance tells applicants to create a first draft themselves and use its assistant to refine it 3, which is a fair description to write into a free-text box even at a company that never asked. The UK Civil Service draws a similar line publicly: refining your own ideas and checking grammar are listed as acceptable, inventing skills or experience is not 4. Both describe a check, not a tool, and that is the register a strong answer borrows.
If the box asks a broader question, how you use AI generally rather than what you did on this one document, what an employer is actually allowed to ask you sets the limit on the other side of that field, since a badly worded question can cross into territory the employer itself isn't supposed to probe.
Not every box on an application means the same thing, either. A yes-or-no toggle near the top of the form, sitting among visa and salary questions, is usually a blanket screen with no bearing on later stages. A more specific question attached to a work-sample upload or a take-home's own instructions is asking about that particular artifact, and deserves an answer scoped to it rather than to your whole job search. Read which kind you're looking at before deciding how much detail to give.
Ask what a badly built box tells you about the process
A disclosure box built well is short, specific, and separate from any question about disability or personal circumstances. A disclosure box built badly asks something vague, offers no room to explain, and sits next to questions that have nothing to do with the work. Read it the way you'd read a badly worded interview question: as information about the employer.
A single vague yes-or-no with no room to explain usually means nobody senior reviewed the form before it shipped, which is worth noting but not worth reading into your own answer. It also helps to know what the answer is likely to be used for. Many employers believe they can spot AI in an application, whether or not they can: in ZipRecruiter's June 2026 survey of 1,000-plus hiring decision-makers, 24% said they can almost always tell, and another 60% said they can sometimes tell 5. That belief is exactly why a box exists at all, and exactly why authorship suspicion alone is a weak basis for a rejection once a candidate has answered plainly. The honest box and the honest answer are the same move from both directions, which is worth knowing when you're the one filling it in.
One more distinction worth carrying past this single field: a checkbox is the honor-system version of a larger question about how an employer actually verifies AI use across a hiring process, and the three approaches employers choose between show where the checkbox sits among them: the cheapest of the three to run, and the least invasive.
A badly built box is also a preview of the rest of the process. An employer that hasn't thought through what it wants from this question, one checkbox, no room, buried in a page of legal boilerplate, often hasn't thought through what it will do with a work sample or an interview either. That isn't a reason to answer less carefully. It's a reason to keep your own record of what you actually did, in case a later stage asks the same question in more detail and you want your answers to agree with each other.
Common questions
What if I'm not sure whether my AI use counts as what the question is asking?
Err toward describing it and let specifics do the work. Say what you drafted with a model and what you checked or wrote yourself, rather than deciding privately that light editing doesn't count. A specific true description of a small amount of AI use reads better than a blanket no that turns out to be technically wrong, and it costs you nothing if the box was only ever checking accuracy.
Does saying yes hurt my chances?
No survey measures that outcome directly, but the attitude evidence points away from panic: half of UK employers in one 2025 survey reported no problem with candidate AI use, and 78% of jobseekers in a separate 2026 survey say they use it at least sometimes, so a yes is closer to ordinary than exceptional. What can hurt you is a yes with no substance behind it, or a no that later turns out false. The honest answer, with a sentence of detail, is the one you can still stand behind at the interview.
What if the form has no AI question at all?
Then nothing on the form requires disclosure, and the decision is yours to make on different grounds: whether the artifact you're submitting is still evidence of something once AI touched it. That is a separate question from what to write in a box that exists, and it deserves its own separate reasoning rather than being folded into this one.
Can an employer actually verify what I put in the box?
Rarely with confidence. AI-text detectors misfire often enough that no serious process should rest on one. The box mostly functions as an attestation, similar to confirming your education or your right to work: something you're on record stating, that the employer can act on if it later turns out to be false, rather than something actively checked at the moment you submit it.
What if my honest answer is that I used AI for most of the writing?
Say that, and describe what you did with the result: which claims you verified, what you changed, what you decided to cut. A heavy draft you reviewed carefully and a heavy draft you pasted unread are different applications even though both start from the same first pass, and the box exists to let you show which one you sent.
References
- 1. 5 trends you need to know from ISE's Recruitment Survey 2025 ise.org.uk Roughly half of employers report no problem with candidate AI use during the recruitment process.
- 2. The 2026 AI in Hiring Report (Section: Candidates turn to AI to keep pace with the process) cdn.prod.website-files.com 78% of candidates say they use AI to tailor their CV or application materials at least some of the time.
- 3. Candidate AI guidance anthropic.com Anthropic's own template for a good answer: draft the first version yourself, then use AI to refine it.
- 4. A candidate's guide to artificial intelligence (AI) in recruitment civil-service-careers.gov.uk A published acceptable/unacceptable split: refining your own ideas is fine, inventing skills or experience is not.
- 5. More Jobs, Higher Bar: The 2026 AI Employer Report ziprecruiter-research.org 24% of hiring decision-makers say they can almost always tell when AI was used in an application, 60% say sometimes.
5 sources, numbered by first appearance. How Olive sources claims
General guidance, not legal advice. Hiring rules differ by state and country and change often; check anything here against your own counsel before you act on it.
Olive assesses how a person works with AI. It does not detect AI-written documents, and it never produces a score, a ranking, or a match percentage for a person. Candidates read the same report the employer reads.