Every hiring team I talk to has an opinion about candidates using AI in interviews. Almost none of them have written it down. That gap — between what we quietly assume and what we've actually decided — is where good candidates get lost and bad decisions get made.
So let's draw the line on purpose.
Start with what you're actually testing
Before you can call anything cheating, you have to know what the exercise is for. A take-home that measures judgment is a different thing from a live problem that measures how someone thinks on their feet — and AI changes each one differently.
A simple framework
- Name the skill the step is meant to measure.
- Decide whether AI assistance changes that measurement — or just speeds it up.
- Tell candidates the rule before they start, not after.
The honest version: most of what we call "cheating" is really just a process we never updated. Fix the process, and the question mostly answers itself.