Nobody has published a credible number for how many of a sports club's followers are real, because almost nobody has checked, account by account. Free "fake follower checker" tools estimate from a sample or from engagement rate — both of which a bot network can fake. The only way to get a real answer is to scan the full following and test every account against real behavioral signals.
Ask a head of digital or a partnerships director what share of their club's social following is fake, and you will get a shrug, a guess, or a number from a free tool nobody has verified. That is not a knock on any one person — it is the honest state of the industry. Follower counts are treated as a fact because they appear on a profile page. They are actually an unaudited claim, and nobody in sports has been asked to audit it until recently, now that sponsors have started asking the question directly.
Why the free bot checkers don't answer this
Type "fake follower checker" into Google and most of what comes back does one of two things: it samples a few hundred followers and extrapolates, or it infers a "quality score" from engagement rate — likes and replies as a share of followers. Both approaches have the same weakness. A sample can miss exactly the kind of farm that was bought in one batch and sits outside the sample. And engagement rate is a metric a bot network can produce on command: accounts scripted to like and reply on a schedule will show up as "engaged," because the tool never asked whether the account behind the like is a person.
Neither approach checks the follower list itself. Neither tells you which specific accounts are real. And neither survives being shown to a sponsor's own analytics team, which is exactly the audience this number increasingly has to satisfy.
What "real" should actually mean
A defensible authenticity check does not guess from a sample — it processes every account in the following and tests each one against signals that are hard to fake at scale. The four signals behind CommunityOS's filter are representative of what a real check looks like:
- Activity floor. Has the account posted, replied, or done anything since it was created, or does it sit dormant the way a purchased or abandoned account does?
- Follower-to-following ratio. Farms and bought accounts skew hard in one direction — following thousands, followed by almost none, or the reverse pattern used to inflate a target account.
- Posting-burst patterns. Dozens of accounts posting the same content within the same second is not a coincidence a human audience produces.
- Linguistic uniformity. Bios and captions that are near-identical across accounts that are supposed to be unrelated strangers.
An account that clears all four is counted as real. One that fails is filtered out before anything else happens — before an engagement rate is computed, before a "top fan" is named, before any number reaches a report. Filtering first is the part almost every tool skips, and it is the part that makes every downstream number trustworthy or not.
The only number we actually have
We are not going to pretend we have a sports number, because we don't. The only production scan CommunityOS has run end to end was on a Web3 project's X following: 78,181 accounts went in, 90.96 percent failed the authenticity checks above, and 5,806 real, rankable accounts came out the other side.
Crypto audiences are about as farmed as an audience gets on X, so that 90.96 percent is closer to a worst case than a typical one. A club's following is inflated too — purchased followers, abandoned promotional accounts, and regional bot farms exist in every sport — but almost certainly at a different rate. Anyone who tells you a specific percentage for a club's following without having scanned it is guessing, us included. The honest position is that the real number for any given team does not exist yet, only the method for finding it.
How to actually find your club's number
The method is the same regardless of whose following is being checked: pull the full follower list rather than a sample, run every account through authenticity checks before scoring anything, and report how many survived and why the rest did not. What comes out the other side is not just a percentage — the accounts that survive get scored and ranked across four archetypes, so the answer to "how many are real" arrives alongside the answer to "which ones matter." The full method, including how the scoring works, is on the methodology page and the engine page.
If the number matters because a sponsor is asking for it, that is a common enough reason to run a scan that it gets its own answer — see how a real, defensible authenticity number holds up in a renewal conversation on the sports page.
What we are asking for
This is also, plainly, a request. We have run this method once, on one Web3 project's audience. What we want next is to run it on one team's following and publish the real number — the actual share of a sports audience that is authentic, measured rather than guessed. If you run a club, a league, or an athlete's account and want to know the number for your own following before anyone else does, that conversation starts at the waitlist.
Quick answers
How many of my club's followers are real?
Nobody knows, and that is the honest starting point. Almost no rights holder has run an account-by-account authenticity check on its own following, so every number in circulation — a follower count, an engagement rate, a free tool's 'quality score' — is a proxy standing in for a measurement that has not happened. The only way to get a real answer is to scan the following and check each account.
Why don't free fake-follower checkers give a real answer?
Most free checkers estimate from a small sample or from engagement rate, not from a check of every follower. Engagement rate is itself farmable — a bot network can like and reply on schedule — so a tool built on it inherits the same inflation it is trying to detect. A credible check processes the full following, applies authenticity signals to every account, and reports how many survived and why the rest did not.
What actually counts as a 'real' account?
An account that clears a set of behavioral checks rather than one that merely exists. CommunityOS runs four signals before anything is scored: an activity floor (has this account done anything since it was created), the follower-to-following ratio (farms and purchased accounts skew hard in one direction), posting-burst patterns (dozens of accounts posting in the same second is a tell, not a coincidence), and linguistic uniformity (near-identical bios and captions across accounts that are supposed to be unrelated people).
Is the 90.96 percent bot rate what our club should expect?
No, and treating it as a sports benchmark would be a mistake. That number came from the first production scan CommunityOS ran, on a Web3 project's X following. Crypto audiences are farmed harder than almost any other category on the platform, so 90.96 percent is closer to a worst case than a typical one. A club's following is inflated too — bought followers and abandoned promotional accounts exist in every sport — but almost certainly at a different rate, and the only way to know which rate is to run the scan.
How do you actually find the real number for a specific team?
Scan the full following rather than a sample, filter with authenticity signals before any scoring happens, and report how many accounts survived. CommunityOS does this as the first stage of every engagement: the filtered, real accounts are then scored across four archetypes and handed back as a ranked list, so the answer arrives as named people worth contacting, not just a percentage.