> it's not hard to detect a bot. So, why can't Twitter just do it?
Presumably a significant part of that is that Twitter cares more about false positives than random third parties do; they're going to get some vicious criticism if they start flagging/closing real accounts as bots. They might also worry more about false negatives, because as soon as they act on bots they'll be accused of bias and only targeting certain positions. (That accusation will hit regardless, but presumably they'd like it to not be true.)
It's easy to whip up a tool that gets lots of true positives, but much harder to get a success rate good enough to use.
Presumably a significant part of that is that Twitter cares more about false positives than random third parties do; they're going to get some vicious criticism if they start flagging/closing real accounts as bots. They might also worry more about false negatives, because as soon as they act on bots they'll be accused of bias and only targeting certain positions. (That accusation will hit regardless, but presumably they'd like it to not be true.)
It's easy to whip up a tool that gets lots of true positives, but much harder to get a success rate good enough to use.