@Sarg0n "close to a DNA test"
Cmon now lets be real.
The 'AI' might work to catch the average goofball who uses a engine with no attempting to mask it.
But; whatever this 'sophisticated AI' is; it does little to nothing at all for those who try to hide it; like using a second device on a VPN to run their engine; while using learning algorithms to simulate human blunders.
Plenty of examples of users who play a crazy number of games to later be marked engine; or youtube videos of users using engines on lichess with no detection.
And there are several threads here and on discord talking about variants like KotH and Antichess not having any actual AI cheat detection and depending wholly on human review.
I realize its good to keep up appearances and all; but lets not pretend that the site is 'cheat proof'
Its cheat resistant; but those who want to cheat; will find a way to do it.
And, the fact that people you report get banned frequently or always shows the inherent bias in the system and problem with human review.
Low ranked people who report a lot; probably have their reports largely ignored. Meanwhile, others have their reports treated like gospel.
Needs a bit of due diligence.
Cheat reports should have absolutely no user name attached to them; for both the reporter and the reported.
I do realize the sheer number of reports that must come in and 'picking and choosing' which ones to review might be only pragmatic.
I think the best thing to keep in mind is:
Lichess does the best that they can.