It's generally much easier to detect a bot than to detect the purpose an account was created for, so I wouldn't be so sure myself.

Also consider that there's also no reason not to allow such accounts. From a business point of view, Twitter exists to sell advertising. Accurately counting the eyeballs (aka mDAU) they have to sell to advertisers is important, but policing the site otherwise makes sense strictly to the extent that failing to do so may drive some of those eyeballs away.

Consider: No real human beings are likely to quit because of fake followers being left alone, but a few vain people might quit if their fake followers were removed. Further, an overly agressive ban wave targeting bots may well remove some bots people actually like which again, on the margin, may drive some real people away. Like any large company, Twitter will inevitably feel the pressure to play it safe, and in this context that means never banning an account that isn't clearly driving real users away - which means that very passive bots that don't do anything are unlikely to be banned.

Of course, none of that gives us any insight into how accurately (or not) Twitter is counting mDAUs.