Are they? Most people will end up using Google or AI or whatever other tool is available to do the job. The tests aren’t for for purpose and that’s the root of the problem, IMO.

These people will then submit PRs with broken code they don't understand, as I have witnessed. You don't know anything about a candidate if you merely witness them repeating what the AI said.

You also don’t know much about them if you rely on online leetcode quizzes, even without AI. That’s the problem. That candidates are using AI is the expected outcome of the enshitification of the interview process.