Straight answer: Twitter’s own 2022 filing to the SEC claimed fewer than 5% of monetisable daily users were bots or spam, Elon Musk’s team argued it was above 11%, and the most cited independent academic study put it between 9% and 15% of all active accounts. None of these numbers are wrong exactly, they’re just measuring different things, and the honest answer is that nobody outside X’s own data team knows the true figure, including X.
The number that started a lawsuit
In May 2022, Elon Musk paused his $44 billion deal to buy Twitter, citing concerns that the company was undercounting bots. Twitter’s SEC filings at the time stated that fewer than 5% of monetisable daily active users, mDAU in their language, were fake or spam accounts. That’s roughly 5 in every 100 users the company was selling to advertisers as real people.
Musk’s legal team disagreed loudly. Court filings from his side argued the real figure could be over 11%, and some of his public statements suggested he believed it was much higher, possibly closer to 20% when you counted dormant, duplicate and automated accounts together. Twitter’s founder-era leadership stood firm on the 5% figure right up until the deal closed in October 2022 for the original price. The bot dispute never went to trial, so the actual disagreement was never resolved in court, it was simply negotiated away.
Bots aside, real accounts notice what you pin, so learn how to highlight brand milestones in a pinned tweet.
That’s worth sitting with for a second. A $44 billion transaction hinged on a number that both sides admitted they couldn’t fully verify.
What independent researchers found before Musk ever got involved
Long before the lawsuit, academics had already been trying to answer this. A 2017 study by Onur Varol and colleagues at Indiana University, published in the ICWSM journal, sampled roughly 14 million Twitter accounts and used machine learning to flag automated behaviour. Their estimate: between 9% and 15% of active accounts showed bot-like patterns. At the time Twitter had around 319 to 336 million monthly active users, so that put the bot population somewhere between 29 million and 50 million accounts.
That study is now nearly a decade old, but it’s still the most rigorous, peer-reviewed number available, and it lines up remarkably closely with the range Musk’s team argued for years later. More recent tracking by companies like Bot Sentinel and Cyabra, which specialise in flagging inauthentic accounts around specific topics, has found much higher concentrations during coordinated campaigns. During politically charged hashtags or crypto pump events, some of these firms have measured bot or troll involvement at 40% to 60% of the total conversation volume, far above the platform-wide average.
So the number moves depending on what you’re looking at. Platform-wide, it’s probably somewhere in the 9% to 15% range on any given day. Inside a specific trending topic or a coordinated push, it can be several times that.
A £1,200 lesson I learned the hard way
Back in 2019 I was setting up an influencer campaign for a finance client. We’d shortlisted an account with just under 94,000 followers and were about to pay 1,200 pounds for a single sponsored thread. Before I signed off the invoice I ran the account through a follower audit tool as a final check, more out of habit than suspicion.
38% of that account’s followers came back flagged as fake, inactive, or bot-pattern accounts. No profile picture, no bio, and a cluster of them created within the same three day window the year before, almost certainly bought in a bulk batch. The account’s real reach was closer to 58,000 people, and the engagement numbers, which had looked healthy at first glance, made a lot more sense once you removed the accounts that could never like or reply to anything.
We didn’t book the campaign. That single check, which took about ten minutes, saved the fee outright. It also changed how I run every influencer deal since, I never take a follower count at face value, and I’d tell anyone doing outbound sales or partnership work through social platforms to build the same habit, whether you’re using dedicated tools or the kind of methods covered in this list of prospecting tools to research a contact before you reach out.
How to check a Twitter or X account’s bot percentage yourself
You don’t need expensive software to get a decent read on whether an account is inflated. Here’s the process I still use before any partnership deal:
- Check the follower to following ratio. Real accounts with genuine influence tend to have far more followers than they follow. A big mismatch in the other direction is a warning sign.
- Look at account creation dates. If you can, check when the account’s most active followers joined. A large cluster created in the same few days is a classic sign of a bought batch.
- Watch posting patterns. Bots often tweet in rapid, evenly spaced bursts at hours no human keeps, or repost identical phrasing across dozens of accounts.
- Check profile completeness. No avatar, no bio, no location and a string of numbers in the handle isn’t proof on its own, but three or four of these together are a strong signal.
- Compare engagement to follower count. A healthy account usually sees 1% to 3% engagement on posts. If someone has 100,000 followers and gets 40 likes, the follower count is doing more work than the audience is.
- Run a proper audit tool. For anything with real money involved, use a dedicated follower audit or influencer vetting platform. If you’re building out influencer relationships regularly, a tool like Heepsy does this vetting as part of the search, which saves you the manual checking I was doing by hand back in 2019.
The part most bot statistics quietly skip over
Here’s the uncomfortable bit. Every number in this post, including Twitter’s own 5%, treats “bot” as a technical category, an account that automatically posts without a human at the keyboard. But that definition is already out of date. A huge and growing share of what functions like bot activity on X today comes from real, verified, human-operated accounts using AI writing tools to generate and schedule content at bot-like volume, or from click farms in places like Bangladesh, India and the Philippines where real people manually operate hundreds of accounts by hand for a few dollars a day. None of those accounts get counted as bots in a company’s SEC filing, because technically a human clicked the button. Functionally, the effect on your engagement numbers, your ad targeting and your sense of who’s listening to you is identical to a bot problem.
There’s also a plainer reason nobody wants to say out loud, which is that platforms have a direct financial incentive to report the bot number as low as possible. Advertisers pay based on user counts and reach estimates. A platform admitting to a bot rate of 20% instead of 5% is admitting to advertisers that a fifth of what they’re paying for doesn’t exist. So any bot percentage a platform publishes about itself should be read as a floor, not a ceiling, and probably a generous one.
Why this matters for your marketing
If you’re spending money on X ads, running influencer campaigns, or building an audience you plan to sell to eventually, the bot question isn’t academic. Vanity metrics like follower count and impressions get inflated by exactly the kind of activity described above, which means your real addressable audience is smaller than your dashboard suggests. This is one of the reasons I push clients to build proper first party lists and think seriously about segmenting their audience rather than chasing raw follower numbers, because a smaller list of verified, engaged real people converts far better than a bloated one padded with dormant and automated accounts. It’s also why measuring influencer partnerships, using something more rigorous than follower count, has become non-negotiable. I’ve written before about how to measure return on influencer marketing ROI, and the bot question sits right at the centre of that, because the wrong influencer with an inflated following will always outperform on paper and underperform in your bank account.
The numbers worth remembering
- Twitter’s 2022 SEC filing: under 5% of monetisable daily active users were bots or spam
- Musk’s lawsuit team’s estimate: above 11%, potentially higher
- Varol et al, Indiana University, 2017: 9% to 15% of active accounts, roughly 29 to 50 million accounts at the time
- During coordinated campaigns around specific hashtags or topics: 40% to 60% bot or troll involvement, per Bot Sentinel and Cyabra tracking
- My own audit on a single influencer account in 2019: 38% fake or inactive followers, on an account with 94,000 followers total
Put those together and a fair working assumption for 2026 is that somewhere between one in ten and one in six accounts you interact with on X is not what it appears to be, and that number climbs sharply the moment a topic starts trending or a coordinated campaign kicks in.
If your question is a different Twitter (X) one, the Twitter (X) guide lists every answer I have written.
Frequently asked questions
What percentage of Twitter accounts are bots according to Twitter itself?
Twitter’s 2022 SEC filing stated that fewer than 5% of monetisable daily active users were bots or spam accounts, a figure the company defended throughout Elon Musk’s lawsuit over the acquisition. Independent researchers have generally put the real figure higher, in the 9% to 15% range.
How can I check if a Twitter or X account has fake followers?
Check the follower to following ratio, look for account creation dates clustered together, watch for rapid or unnatural posting patterns, check whether profiles have photos and bios, and compare engagement rate to follower count, a healthy account typically sees 1% to 3% engagement. Dedicated follower audit tools or influencer vetting platforms will do most of this automatically.
Are bot percentages higher during trending topics?
Yes. Platform-wide bot activity sits somewhere around 9% to 15% on an average day, but during coordinated campaigns around political hashtags, crypto promotions
