The basic measure is “accuracy”: the percentage of moves that are the best move for that position. Computers are, as you would expect, inhumanly good on this measure. Better than the best human players, who will have “inaccuracies” even in their best-played games. It makes spotting cheaters pretty reliable and easy. Niemann has been busted for cheating online repeatedly, and analysis of his in-person play bears the same hallmarks.

BBC article summarizing some of the suspicions and evidence or lack thereof.

https://www.bbc.co.uk/news/uk-63140246

The main part of the report is only 20 pages. And there are a lot of figures.

This is "a whole lot of evidence". If he had, for example, a 10% chance to show up suspicious in any one of those charts that's one thing. But to be highlighted as the most suspicious in each is extraordinary evidence. How do you explain a greater than 10% drop in skill when a 15 minute TV delay was put in place? Or his ability to make incredibly complex, perfect moves in seconds? Or his continual, nearly unstoppable strength rating growth, you know, except for the two natural plateaus in rating where most players never continue to grow firmly in the middle of his growth curve?

If that report doesn't convince you, I doubt anything will.

If he always did what the engine tells him to do then the evidence would be overwhelming, yes. Hans is surely aware of this and avoids doing so because he wants to get away with it.