Accuracy score has been changed a few years ago. It is usually significantly higher today. The most drastic examples are lower accuracy games a few years ago. I have some games in 20 range back then. When I do accuracy check today, they are around 50-60 for instance.
After the change, I had an unrated game against sub 1000 player that lasted around 10 moves where he blundered 3 times and got around 70 accuracy.
I play relatively longer games and I feel that the only time I get lower accuracy is when the game gets sharper.
The same goes for estimated rating. By its estimation, I am usually over 2 000 rated.
All of these are made to make us feel good about our game, even if the reality is a bit different. The point is of course to give us a better experience and to have more people with memberships. I don't like it, but it is understandable.
In practice, people are now a lot more suspicious.
Sometimes it is justified, cheating exists, it is undeniable, but the lower the rating, these people are easier to detect.
Chess masters are potentially much better cheaters because they know more about the game and they know better what seems more natural, which is why it takes longer to catch them.
Sub 500 player cheating is usually like a todler in a museum claiming he painted Mona Lisa. That is usually how obvious low level cheating is.
Are you seriously quoting an "AI Overview"? Wow.
Those things written there have almost nothing to do with how chess.com really uses "accuracy". It doesn't mention the fact that chess.com modifies the accuracy score based on the players' rating, giving lower rated players higher scores no matter what.