Or it could be that Twitter cares about fake account impact mainly in terms of user experience. If a fake bot likes a tweet from another fake bot, does any human actually care?
I used to have a much bigger problem on Twitter with fake accounts and bot likes than I did, but despite my decades of willingness to bitch about spam, I have to concede that they've gotten better lately.
There's also a user benefit to letting bots run when they're not hurting anyone: you don't give hints to the spammers on how you're finding and nuking them. Indeed, if you identify them but don't block them, you can get a lot of data on what is spam. For example, on my mail server, I noticed I was getting a lot of dictionary attacks. So I took the hundred most common first names not in use on my domain and fed them all into the spam training system. That means odds are very good that my spam trainer will have seen a piece of spam before they try my actual account.
You get to see a flood of spambots if you make the mistake of clicking any of the trending hashtags, but otherwise yeah it's a non-issue for the vast majority of Twitter usage.
I used to have a much bigger problem on Twitter with fake accounts and bot likes than I did, but despite my decades of willingness to bitch about spam, I have to concede that they've gotten better lately.
There's also a user benefit to letting bots run when they're not hurting anyone: you don't give hints to the spammers on how you're finding and nuking them. Indeed, if you identify them but don't block them, you can get a lot of data on what is spam. For example, on my mail server, I noticed I was getting a lot of dictionary attacks. So I took the hundred most common first names not in use on my domain and fed them all into the spam training system. That means odds are very good that my spam trainer will have seen a piece of spam before they try my actual account.