The tl;dr is that all improvement in the quality of play comes before move 60, when humans can mimic memorised AI policies. Play after move 60, in the pivotal parts of the game, shows no improvement. For me to think there’s any meaningful change in human play from pre-AI times, I would have to be convinced that players understand the AI moves they copy well enough to keep a heightened level when they go off-policy after the opening. There is no evidence of this.
An alternate hypothesis is that the first 60 moves are where AI has generated the most human-accessible insights. We already understood endgames pretty well before AI (afaik 19th-century masters would play almost perfect endgames when they had multiple days per move), so maybe there just isn’t much room for improvement. Human professionals beat katago-micro (~30 playouts), so it’s possible that the middlegame intuition of humans is better than that of AIs but the AIs have much more search. In the opening, however, AIs completely upended traditional opening theory in a way that humans could understand. Even if you can’t continue to match the AI’s strength in the midgame, you can still build up a lead of a couple of points. This won’t give a decisive win, bit at the professional level, komi is certainly a big deal, so I would imagine that 2 or 3 points of advantage in the opening is quite useful. I also find it hard to believe that players would abandon joseki supported by hundreds of years of tradition if it wasn’t even helping them win.
I also think it could be interesting to see an analysis of the number and strength of top amateurs. Even if AI isn’t that helpful if you’re already an insei training with top pros, it could still make it easier to reach high amateur dan ranks without as much access to strong humans. Personally I have gotten nonzero benefit from reviewing games with AI. If you only play with opponents around your level and don’t have a stronger player to review your games, it can be easy to get stuck in blind spots and not even know what you’re doing wrong. E.g. sometimes I think I messed up a tricky defense, but really the AI shows me I overplayed way earlier and it was inevitable.
I don’t think there is anyone using this to counter cheaters, but people do it for fun, and there is definitely more than one example.
Here is a video where Nick Sibicky beats katago with 10k playouts.
Here is a game on ogs where a player captures the entire board against katago micro.