Assessment design
When an Assessment Is Silent on AI, Ask Before You Start
Silence isn't permission and it isn't a ban. Send one message to the recruiter before you open the assessment: is an AI assistant allowed on this task, and do you want me to say what I used. The message takes a line to write, and the reply tells you something about the employer. If nobody answers before the deadline, work in a way you can walk someone through, keep a note of what you used, and say so when you submit.
The takeTwo opposite pieces of advice circulate, and both are bad for the same reason. Using it quietly because everyone does, and never touching it in case you get caught, both treat the employer as unreachable. A recruiter who cannot answer a one-line question about their own assessment has told you something worth knowing before you spend three hours on it. Asking costs a message. Guessing can cost the round, and you may never find out which way you guessed wrong.
Where Olive fits
Open a role and see what the work shows
If the assessment you're sent is Olive, there is no silence to resolve, because the AI assistant is part of the work: a person then writes or edits six findings on what happened, in words rather than a score. You're granted the identical copy the employer reads, free.
Assess your shortlistAsk Before You Open the Assessment
One message to the recruiter, sent before you open the assessment, resolves this faster than any amount of reading. Two questions fit in a line: is an AI assistant allowed on this task, and if it is, do you want me to say what I used. Send it to whoever scheduled the round. A reply can be a single sentence.
Asking is not a confession.
One survey puts a rough size on the gap. In the Institute of Student Employers' 2025 recruitment survey, 45% of the UK early-careers employers surveyed had given applicants no guidance at all on when AI use is or isn't appropriate 1. ISE's membership is self-selected and the full report is member-gated, so the sample behind that figure isn't public. Read it as a rough shape rather than a rate.
The silence may be nothing more than an oversight.
What comes back is information either way. An employer who answers in a sentence has thought about the round. One who says the policy is still being worked out has told you the rule may land differently on different candidates, which is worth knowing before you invest an evening. Employers are being told to set the AI rule stage by stage and write it where candidates read it, so a missing rule reads more plausibly as a gap in their process than as a trap set for you.
Why the Rules Genuinely Vary
No default is safe to infer, because the rule changes with the platform, the test and the employer, and it has changed over time at the same vendor. A forum thread about what one company allowed last year tells you almost nothing about the assessment sitting in your inbox today. That is the case for asking rather than for picking a side and hoping.
HackerRank states the per-test framing on its Screen product page: a test designed to allow AI is fair for everyone to use it on, and a test that explicitly instructs otherwise will have AI use flagged as suspicious activity 2. Same platform, opposite rules, set by whoever configured the test. The page lists the signals an employer can switch on without saying which are on by default 2, so the instructions you were sent stay the only place the rule for your test is visible to you.
HackerRank is describing its own product there, and no accuracy figures are published beside those signals.
A vendor's own rule can move too. Karat's candidate page, dated 9 February 2023, let candidates look up documentation during the coding portion and use ChatGPT for that if they told their Interview Engineer first, while ruling out using it to generate an approach or a solution 3.
On 10 December 2025 Karat launched NextGen Interviews, where engineers work a multi-file project with an integrated AI assistant while an Interview Engineer probes their reasoning live 4. NextGen is one product line, not a statement that every Karat interview now allows an assistant, which is exactly why the answer has to come from the employer running your round.
Other employers write the opposite default. IT Pro reported in March 2025, relaying internal documents seen by Business Insider, that Amazon's recruiter guidelines tell candidates not to use generative AI during an interview unless explicitly permitted, and warn that failing to comply may result in disqualification from the recruitment process 5.
That wording reaches you third-hand, IT Pro relaying Business Insider, and the documents behind it were not published. Amazon's on-record position, given to CNBC and NBC News by a company spokesperson the same month, is that candidates acknowledge they will not use unauthorized tools such as generative AI during the interview or assessment process 5.
Where a guideline reads "unless explicitly permitted," silence resolves to no.
What If Nobody Replies Before the Deadline?
Do the work in a way you can walk someone through, and say what you used when you submit. That position doesn't depend on guessing which answer applied: whatever the rule turns out to be, you'll have hidden nothing. Keep your own short record while you work: the tool, what you asked it for, and what you changed afterwards. Two or three lines is enough.
Declaring is not an apology. Karat's 2023 page asked candidates to tell the Interview Engineer before using ChatGPT to look up documentation, which is a vendor writing a close cousin of this move into its own live-interview instructions 3. A note at the top of the README costs a sentence: which parts you wrote, where an assistant helped, and what you tested. Written plainly it reads as method, and a reviewer who has it knows what to ask you about first.
One sentence, written while the work is fresh.
Assume someone may ask you to walk through it. HackerRank's Screen page also lists full session replay among the signals an employer can switch on 2. A follow-up conversation may come anyway, with or without a platform feature behind it. What actually happens when a take-home says no AI covers the same ground where the rule is stated rather than missing.
The question to prepare for is some version of: talk me through this.
Include the note even when the deadline arrives with no reply. An unanswered question you can point to is a different position from a silence you never tried to break, and the record of asking sits in your sent mail where you can find it.
What Makes a Submission Defensible Either Way
Defensible means you can account for every decision in it, in your own words, without the tab open. Whether an assistant helped is the second question, not the first. A defensible submission is one where you chose the approach for a reason you can state, tested it against a case you picked, and know which parts you'd change first.
- Say what the constraint was before you say what the code does.
- Name one case you tested that the instructions didn't hand you.
- Point at the part you trust least, and say why.
None of that depends on whether an assistant was open.
If the answer comes back that AI is allowed, the bar moves rather than drops. When AI is allowed in the coding round, judgment is what gets read describes what an interviewer watches for once finishing stops being the test.
The survey's other figures are worth carrying into all of this. In the same ISE survey, 49% of employers said they had no problem with candidates using AI during recruitment, and 10% had banned it or introduced technical measures to prevent it 1.
Those are attitudes reported by employers in a survey rather than policies you can rely on, and they cover UK early-careers employers only. Nothing in them says what the employer holding your assessment decided.
They are still a reason not to read silence as a threat.
Employers are working through whether to let candidates use AI in the interview or ban it, and that guidance tells them to put the answer in the invitation. Reading that page also shows what a considered rule looks like, which makes a vague one easier to spot. Until the employer holding your assessment does that, the message you send is doing the work for them.
Common questions
Does silence mean AI is allowed?
No, and it doesn't mean banned either. An assessment that says nothing carries no rule you can read, so the move is to ask the recruiter before you open it: is an assistant allowed on this task, and do you want it declared. Rules vary by platform, by test and by employer, and they have changed over time at the same vendor, so there is no default worth inferring.
What exactly should I write to the recruiter?
Two sentences to whoever scheduled the round. Name the assessment, then ask whether an AI assistant is allowed on it and whether they want you to say what you used. Nothing else is needed. If the deadline is close, say when you plan to start, so a reply arrives while it still changes what you do.
Will asking make me look like I was planning to use AI anyway?
Asking about the rules of an assessment is the same kind of question as asking whether it is timed or which language to write in. A recruiter who reads it as an admission has told you something about the process you were about to enter. The alternative is guessing on a rule that varies by employer, and a wrong guess costs more than the message did.
Nobody replied and the deadline is tonight. What do I do?
Do the work in a way you can walk someone through, keep a short note of what you used, and put that note in the submission itself: which parts you wrote, where an assistant helped, and what you tested. That position doesn't rely on guessing whether the answer would have been yes or no, because nothing in it is hidden. A submission you can explain end to end is what a follow-up conversation is looking for.
Should I declare AI use if nobody asked?
Yes, in one line, when the instructions are silent. It costs one line if AI was allowed, and it is the difference between a disclosed choice and a hidden one if it wasn't. Karat's 2023 candidate page asked for something similar in its own live interviews: tell the Interview Engineer before using ChatGPT to look up documentation during the coding portion 3. Keep it factual: the tool, what it did, what you checked.
References
- 1. 5 trends you need to know from ISE's Recruitment Survey 2025 ise.org.uk Supports the article's base rate: 45% of surveyed employers had given applicants no guidance on appropriate AI use, 49% had no problem with candidates using AI in recruitment, and 10% had banned it or introduced technical measures against it. UK early-careers employers in a self-selected membership body; the underlying report is member-gated, so no sample size is public.
- 2. HackerRank Screen (product page) hackerrank.com Supports the per-test framing quoted in the article (a test designed to allow AI versus one that explicitly instructs otherwise, where AI use is flagged as suspicious activity) and the mention of full session replay as a signal an employer can switch on. A vendor page with no accuracy figures published.
- 3. Can I use ChatGPT for my technical interview? karat.com Supports the 9 February 2023 Karat position the article describes: documentation lookup with ChatGPT is allowed if the Interview Engineer is told first, and generating an approach or a solution is not. Also supports the article's point that declaring a tool before using it is a move a vendor already asks for.
- 4. Karat launches NextGen Interviews: the first human-led, AI-enabled talent evaluation solution karat.com Supports the 10 December 2025 launch the article uses to show one vendor's rule moving: engineers work a multi-file project with an integrated AI assistant while a live Interview Engineer probes reasoning. A press release about one product line, not a claim about every Karat interview.
- 5. Amazon banned candidates from using AI tools during interviews itpro.com Supports the article's example of an employer whose default is the opposite of permission: recruiter guidelines telling candidates not to use generative AI unless explicitly permitted, with non-compliance warned as possibly resulting in disqualification, plus Amazon's on-record spokesperson statement about unauthorized tools. The disqualification wording is third-hand: IT Pro relaying Business Insider's report of internal documents.
5 sources, numbered by first appearance. How Olive sources claims
General guidance for hiring teams. What works at one company and one volume may not transfer to yours.
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.