How Do You Spot an Influencer With Fake Followers?

Fake followers are usually discussed as a scam to be caught, which gets the shape of the problem wrong. Some fake following is involuntary, most creators have some, and the useful question is not whether a creator is clean but where your threshold sits and what you are paying for.

In the Influencer Marketing Hub’s 2026 benchmark survey, only 10.9% of brands reported running into none of the category’s fraud and quality problems. Nine in ten met at least one, which tells you what the base rate looks like before you have checked anybody.

How common are fake followers, really?

Fake followers are common enough that you should plan around the base rate rather than the exception. According to the Influencer Marketing Hub’s Influencer Marketing Benchmark Report 2026, only 10.9% of brands reported running into none of the category’s fraud or quality problems, so roughly nine in ten met at least one. The realistic goal is screening out the severe cases, not finding a pristine audience.

Fake or bot followers are the dominant complaint by a wide margin. In the same survey they accounted for 56.5% of all fraud and quality issues reported, against 10.6% for inauthentic or templated comments and 10.2% for fake or purchased engagement. The audience gets faked far more often than the engagement layered on top of it.

What those numbers do not tell you is how much of it is deliberate. A brand reporting a fake-follower problem has found fakes, not proof the creator bought them, and the distinction decides whether you renegotiate or walk.

The platforms are removing accounts at a scale that reframes the problem. According to Meta’s own Community Standards Enforcement Report, Meta actioned 1.1 billion fake accounts on Facebook in Q4 2025. In May 2026, an Instagram sweep removed millions of accounts in a matter of hours and visible follower counts fell across the creator economy.

That last point is the one to take into a negotiation. A creator whose count dropped sharply in a purge was carrying fakes, whether or not they bought them, and the current number is more trustworthy than the one in the media kit.

What does a fake follower actually look like in the data?

A fake follower leaves a shape in the data rather than a single tell, and the shape of a bought following is more reliable evidence than any individual number. Four patterns do most of the identifying, and each one is visible without a paid tool.

The growth curve is the clearest. Real accounts grow in uneven steps that line up with specific posts, and each step has a post you can go and look at. A bought following shows vertical jumps with nothing underneath them, or a suspiciously straight line of steady daily additions that no organic account produces.

Audience geography is the second. A creator posting in English about a US product whose audience is concentrated in markets with no connection to either is the most common signature of cheap follower packages, which are sold from wherever they are cheapest to supply.

Comment text is the third and takes ten minutes. Read thirty comments across three posts. Generic praise, emoji strings and the same short phrases appearing under unrelated posts indicate an engagement service rather than an audience. Real comments reference the specific thing in the specific video.

The fourth is the ratio of views to followers on recent posts, read against that creator’s own history. A creator whose views have stayed flat while followers climbed has added people who do not watch. That comparison against their own past is the honest version of the engagement check, and it is more useful than measuring them against a published benchmark.

What you look at	A real audience	A bought one
Follower growth	Uneven steps, each with a post behind it	Vertical jumps, or a flat daily trickle
Audience geography	Matches the creator’s language and market	Concentrated in unrelated cheap-supply markets
Comments	Reference the specific thing in the video	Generic praise and emoji, repeated across posts
Views against followers	Move together over time	Followers climb, views stay flat
Engagement mix	Likes and comments in proportion	Heavily skewed to likes

How do you check an influencer’s followers before you pay?

Check in a fixed order so the cheap tests come first, and stop as soon as something fails. The whole sequence takes about twenty minutes per creator and catches most of what matters.

Start with the public profile. Scroll the recent posts and compare view counts, like counts and comment counts for internal consistency. Posts with high likes and almost no comments, or comments that are all one line long, are the pattern to notice.

Then ask for native analytics, as a screen recording rather than a screenshot. A screen recording of the creator scrolling their own insights is hard to fake and takes them two minutes. Ask specifically for audience country breakdown, audience age, reach split between followers and non-followers, and the last 30 days of follower growth.

Then compare what they sent against what you saw publicly. The two should agree. A media kit claiming an audience that is mostly US against an insights screen showing something different is the end of the conversation, and this is the step most brands skip.

Run a paid audit last, and only on creators you are about to commit real money to. Audit tools build a composite score from growth patterns, engagement authenticity and audience composition, which is useful triage. It is worth remembering that the firms publishing the industry fraud statistics also sell fraud detection, which argues for reading their numbers carefully rather than taking the score as a verdict.

For creator sourcing more broadly, including where to find people worth running this check on, see how to find and vet UGC creators.

Can a high engagement rate also be a warning sign?

A high engagement rate can absolutely be a warning sign, and this is the check most brands get backwards. Engagement that is far above what a creator’s size and category normally produce is as suspicious as engagement far below it, because purchased engagement is the easier fraud to buy and the harder one to notice.

The giveaway is the composition rather than the level. Bought engagement skews heavily towards likes, because likes are cheapest. A creator with a very high like rate and almost no comments, or comments that do not respond to the content, is showing the signature of a service rather than an audience.

Timing is the second giveaway. Real engagement arrives in a decaying curve over hours and days. Purchased engagement arrives in a block, usually soon after posting, and then stops. If a creator will show you a screen recording of a post’s insights, the shape of that curve is visible.

Comparing against benchmarks is the weakest version of this check, because engagement rate falls as accounts grow, differs sharply between platforms, and drops when someone posts more often. A creator sitting below a published average may simply be bigger or more prolific than the average. What matters is their own trend. The benchmark numbers themselves are in what is a good engagement rate for influencers.

Treat engagement as one input and never as the decision. A creator with modest engagement whose audience buys things is worth more than a creator with excellent engagement whose audience does not, which is a different measurement problem entirely.

What should you do when a creator has some fake followers?

With some fake followers, renegotiate the basis of the deal rather than walking away. A following with a suspicious share in the mid teens is normal, and the right response is to stop paying for reach and start paying for something fraud cannot inflate.

Three restructures work. Pay a content fee plus usage rights instead of a sponsorship fee, so you are buying the video rather than the audience. Put part of the fee behind a result you can verify, such as link clicks or sales. Or reduce the fee in proportion to the audience share that is actually in your market, which is a conversation most professional creators will have.

The licensing route is the cleanest and the most underused. If the plan is to run the creator’s video as a paid ad, the creator’s own audience is close to irrelevant and the only thing that matters is whether the content performs. Fraud exposure drops to almost nothing, and the content usually outperforms the organic post anyway. Rates for that structure are in how much UGC usage rights cost.

Walk away when the evidence is of deliberate purchase rather than accumulation. A vertical spike in the growth curve, an audience concentrated in markets with no relation to the creator’s content, and a refusal to screen record their insights together make a pattern worth declining on, whatever the content looks like.

Keep a record either way. The creators you declined and why is a useful internal asset, because the same accounts are pitched to the same brands repeatedly and re-running the check from scratch each time is wasted work.

How do you write the contract so this is the creator’s problem too?

Write it so that audience quality is warranted, verifiable and partly paid for on results, which moves the risk to the party who actually knows the answer. The creator knows whether they bought followers. You are guessing.

Four clauses carry most of the weight. A warranty that the creator has not purchased followers, likes, comments or views. A right to request native analytics during the engagement, not just before it. A performance floor tied to a metric fraud cannot manufacture, meaning clicks, signups or sales rather than impressions. And a staged payment, with a meaningful share held until the agreed deliverables and results are confirmed.

Keep the floor realistic and keep it simple. A floor set at an optimistic number gets negotiated out or produces a dispute, while a floor set just below the creator’s own recent averages is easy for an honest creator to accept and immediately uncomfortable for a dishonest one.

Avoid the clause that asks the creator to guarantee sales. It shifts a risk they cannot control, good creators decline it, and you end up selecting for the people with no better options. Guarantee the inputs, share the upside on the outputs.

The rest of the terms, including usage rights, exclusivity and approval windows, are covered in what should be in an influencer contract.

The short version

Some fake following is normal. According to the Influencer Marketing Hub’s 2026 benchmark survey, only 10.9% of brands had run into none of the category’s fraud or quality problems, so screening is about thresholds rather than purity. Under a tenth is noise, a tenth to a quarter deserves a conversation, over a quarter is a reason to decline unless everything else is strong.

Check the shape, not a single number. Look for vertical jumps in the follower curve with no post behind them, audience geography that does not match the creator’s market, generic repeated comment text, and views that have stayed flat while followers rose.

Ask for a screen recording of native insights rather than a screenshot, and check it agrees with what the public profile shows. Very high engagement deserves the same suspicion as very low engagement, because likes are the cheapest thing to buy.

Then make the structure do the work. Buy content and usage rights rather than reach where you can, since a licensed video run as a paid ad barely depends on the creator’s audience at all.

Warrant audience quality in the contract, hold part of the fee against a result fraud cannot fake, and expect follower counts to fall during platform purges: according to its own enforcement report, Meta actioned 1.1 billion fake accounts on Facebook in a single quarter, and the Instagram sweep in May 2026 took millions more out in hours.