I know a few people who do Twitter bot detection as sort of a hobby, and I would say you have to be skeptic of these numbers from both directions. Obvious bots

I know a few people who do Twitter bot detection as sort of a hobby, and I would say you have to be skeptic of these numbers from both directions. Obvious bots are, well, obvious. Novice or careless implementations lead to accounts that tweet at the same time or based on the same trigger words. Tweets made on a routine basis are the easiest to detect (e.g. Tweets at 5PM CST every day) But things go much deeper. There are even companies that now develop bots tools essentially, increasing the randomness or human-like qualities of bots. These are much harder to detect with the standard methods. There are other proxies, like not getting organic replies or having no friends. But the sophisticated bot networks have already figured this out. You have to realize that sometimes human labor even goes into the effort, where someone will maintain hundreds of fake accounts, and attempt to make them look organic. From these points, it's easy to underestimate the number of actual bot accounts. On the other hand, these same methods don't usually control well for a certain type of Twitter user. There are actually some really obsessive twitter users who will spend 12+ hours a day active. They typically produce a ton of content, and a lot of replies. I notice this on celebrity/political accounts especially. (Apparently people are calling this the 'reply-guy' now) Those accounts are susceptible to appearing as bots because they deviate far from what one might imagine the 'average' Twitter user looks like. Speaking anecdotally, I think this relates closest to politics, and hot topic issues (Hence the topic in this research). Take a Tweet from the US President for example. You will notice the same few dozen accounts reply to every Tweet of his. Some may be bots, but others are most definitely human. But these often get classified as bots. If I had to make a guess, I would say the bot detection models fail more often to account for the true number of bots. At this point, it's an arms race. Obvious bots are becoming increasingly less obvious. Overall, I think bots are only going to get better, and it worries me for the fate of social networks. The thing most frightening is finding an obvious bot and seeing a large number of organic users interacting with it as if it were real. Edit: One more thing I forgot to mention. An increasing number of bots are getting verified now. I have no idea how the verification process works, but apparently the botters have figured out how to game it successfully.