I've conducted interviews where the candidate asked if he could use google to try to get an answer. I often say "sure". If a guy can read an explanation out of context, understand it in a way he can explain it using his own words, and reason about corner cases in a couple of minutes, he's hired. The same goes with AI; canned responses work when you ask canned questions, not so much on open-ended ones.
That's missing the point. The goal is to have a level playing field for the interview.
If your interview format allows people to use outside help but only if they think to ask, that's hardly a level playing field. You're testing the candidate's willingness to ask. In most interview formats it would not be acceptable to Google the answer, so most people won't ask.
If you have an interview format that allows Googling, you should mention that at the start. Not leave it as a secret for people to discover.
The questions dont require google; but what do you do when you don't know a specific thing? You search for it.
The notion that a candidate must remember the name of a thing or a specific algorithm is just ridiculous. When was the last time you implemented some fancy sorting or tree traversal algorithm from memory?
and if a guy thinks he's able to parse that amount of information in less than a minute, why should I refuse it? The end goal is to hire problem solvers, people with analytical thinking and capable of learning autonomously.
In most companies, the development process is collaborative - spikes, code reviews, informal meetings; why would you evaluate a candidate for such a team solely on what narrow knowledge he brings to the table when the power is down?
