X bots are about 9% to 15% of active accounts

ByJames McCallef
x bots: Editorial image suggesting social media bots and automated accounts

The best public estimate is that X bots make up roughly 9% to 15% of active Twitter accounts, based on a widely cited 2017 study from researchers at Indiana University and the University of Southern California that estimated 9%-15% of active accounts are bots (arXiv). A commonly repeated translation of the high end is about 48 million accounts, which Axios got by applying 15% to Twitter’s then-reported 319 million monthly active users (Axios).

That is the answer most readers want, and it is also about as precise as the public evidence allows. There is no single reliable live count because bot is not one thing, public researchers do not get full internal platform data, and different studies measure different targets: accounts, activity, links shared, or spam behavior.

How many bots are on X?

The most defensible estimate still comes from the 2017 paper Online Human-Bot Interactions: Detection, Estimation, and Characterization, which concluded that between 9% and 15% of active Twitter accounts are bots (arXiv). That is old, but it remains one of the clearest independent account-level estimates in public.

The “48 million Twitter bots” figure is not a separate measurement. It is a back-of-the-envelope conversion of that 15% estimate using Twitter’s 319 million monthly active users at the time, which yields roughly 47.85 million accounts (Axios). Useful shorthand, yes. A fresh census of X bots, no.

Pew Research Center found something different, and often misread: 66% of tweeted links to popular websites were posted by suspected bots, and the 500 most active suspected bot accounts produced 22% of tweeted links to popular news and current-events sites (Pew Research Center; Pew Data Labs). That shows bots can dominate activity without necessarily being most accounts. A small number of loud accounts can make the platform feel much more synthetic than the raw account share suggests.

Why there is no single count

The first problem is definition. A spam bot, a news-posting automation account, and a customer-service auto-responder are all automated in some sense, but they are not equally malicious or equally likely to be counted in the same study. The arXiv paper estimated active accounts (arXiv); Pew looked at tweeted links shared by suspected bots (Pew Research Center). Those are related questions, not interchangeable ones.

The second problem is access. Outside researchers can sample public behavior, but the platform has far more signals: IP patterns, phone verification, device fingerprints, suspension histories, and internal abuse tooling. Without those, public estimates are necessarily probabilistic rather than authoritative. That does not make them useless; it just means they come as ranges, not a neat dashboard number.

The third problem is that the target moves. Bot operators adapt to detection methods, and platform policies change over time. A bot estimate from 2017 is not a live reading for 2026. It is still the best-cited independent benchmark, but not a permanent truth engraved into the API.

What the best estimates say

If you want a one-line answer, use this: independent research suggests about 9%-15% of active X/Twitter accounts are bots, with “about 48 million” as a rough high-end translation rather than a direct count (arXiv; Axios).

The important nuance is this:

  • 9%-15% refers to estimated active accounts in a major academic study (arXiv).
  • About 48 million is a media conversion of the 15% figure using Twitter’s user base at the time (Axios).
  • 66% of tweeted links refers to content activity measured by Pew, not the share of all accounts (Pew Research Center).
  • 22% of links to popular news/current-events sites came from just 500 most active suspected bot accounts, which is a good reminder that concentration matters as much as prevalence (Pew Data Labs).

In other words, asking “how many bots are on X?” sounds like a simple inventory question. It is really three questions: how many automated accounts exist, how many are active, and how much of the visible conversation they drive. Public research answers the second and third better than the first.

Key Takeaways

  • Independent researchers estimated that 9% to 15% of active Twitter accounts are bots in a widely cited 2017 study (arXiv).
  • The often-repeated 48 million figure is a calculation from that estimate, not a separate direct measurement (Axios).
  • Pew found that suspected bots posted 66% of tweeted links to popular websites, which measures activity share rather than account share (Pew Research Center).
  • There is no single reliable public bot count because the number changes with the definition of bot, the sample, and the data researchers can access.

Further Reading