It's problematic to identify accounts as bots. Many accounts are partially posting from a bot and partially posting from a human, so what are they? Better to do that at the level of individual posts then the user would see right there that something was posted by the api, and it'd be harder for political spambots for example to flood threads, you could have options to turn off api posts for a thread. This would drive some bots to use selenium etc (perhaps some do already), but I think would be more effective than just marking accounts somehow as 'bot' with fallible detection methods.
Also most of the problems on twitter are caused by humans, not bots. Humans can be cruel, harass, bombard victims with messages, threaten them with death, rape etc (this is regularly happening right now to women on twitter), and those are the more problematic interactions. Yes sometimes spambots liking posts or replying to posts are annoying, or the ones that DM you are pretty bad, but they're not as annoying as humans at their worst.
