For interviewers · take-home review
Upload candidate submissions and get a per-file report that shows which answers deserve deeper questions. Use it to focus the interview, not to auto-reject a candidate.
A screening signal is a reason to ask more, not a reason to reject.
Strong candidates can write clean code quickly. Weak candidates can also paste polished output from an AI tool. The useful question is not "is this definitely AI?" It is "which submissions should get a deeper code walkthrough?"
The report highlights unusual patterns and gives your team a place to start. Bring the flagged snippets into the interview and ask the candidate to explain the design, the tradeoffs, and the parts they would change.
A Hacker News launch that claimed high accuracy drew comments showing both false positives and missed AI code.
If you do not have Codility or HackerRank in the stack, you still need a consistent way to decide where to dig. Use a shared report so interviewers ask about the same snippets instead of trading gut feelings.
No. Use it to choose follow-up questions. A detector score should not be the reason to reject a candidate.
Ask the candidate to explain the flagged code, change one requirement, and discuss tradeoffs. People who wrote the code usually handle that well.
That may be normal in your process or not allowed at all. Set the policy first, then use the report to guide the conversation.
No. It does not compare code against other candidates or public sources. It screens for signals associated with AI-generated code.
Screen one submission free, then use batch review when take-home volume grows.