1 Million Bot Checks Reveal When We Really Suspect Fake Accounts

Most studies of fake accounts look at the bots themselves. This one flips the camera around and studies us — the people doing the suspecting. The researcher analysed server logs from Botometer, a popular free tool where anyone can paste a Twitter handle and get a bot-likelihood score. Every one of the 1 million-plus checks made between 2020 and 2023 was treated as a small behavioural trace: a moment when a real person thought, "something about this account feels off."

What triggered those moments? Collective suspicion surged during platform-wide crises, rising most sharply around the 2022 public feud between Elon Musk and Twitter over how many bots were on the site. The accounts people checked were older, posted far more often, had bigger follower counts, and disproportionately pushed promotional, political, and cryptocurrency content. In other words, this wasn't random paranoia — it concentrated on the loudest, most agenda-driven accounts on the timeline.

And the crowd was largely correct. Accounts that attracted collective suspicion scored higher on the bot-detection scale and were more likely to be suspended later. That is a striking finding: ordinary users, armed with nothing but a hunch and a free tool, were flagging the same accounts the platforms eventually acted against. The researcher describes this as a mass, distributed form of platform auditing — millions of small citizen investigations adding up to something meaningful.

The study also looked at how people argue. When defending a human, they leaned on first-person identity claims ("I'm a real person, here's my life"). When accusing a suspected bot, they cited evidence: posting rate, political content, and coordination across accounts. The takeaway: public bot-suspicion isn't noise. It tracks real signals, and tools like this turn everyday distrust into usable data.