Spotting fake followers: the signals that work

Fake followers are the oldest problem in influencer marketing, and they haven’t gone away. A creator with 100,000 followers, a fifth of them bots, is a creator with 80,000 followers charging for 100,000. Fake engagement is worse: it makes a weak creator look like a strong one.

Most influencer databases now sell a fraud score that does the first pass for you. Those scores are useful, but they’re estimates built from a sample, made by methods vendors only partly publish. The best approach is a score to screen a long list, then a few minutes of your own checks on anyone you’re about to pay.

What counts as fake

“Fake followers” covers a few different things, and the tools treat them differently.

• Bots and bought followers. Accounts created to follow, often in bulk, with no real person behind them. This is what most people mean.
• Mass followers. Real accounts that follow thousands of others, so they rarely see any one creator’s posts. HypeAuditor defines them as real accounts following more than 1,500 others. They aren’t fraud, but they aren’t much of an audience either.
• Inactive accounts. Real people who stopped using the app. Every creator has some; older accounts have more.
• Fake engagement. Bought likes, comments and views, and engagement pods, where groups of creators agree to like and comment on each other’s posts. This inflates engagement rate rather than follower count, and it’s harder to spot.

The distinction matters when you read a score. A creator can have a lot of mass followers and inactive accounts without ever having bought anything, and a creator with a clean follower list can still be buying comments.

What fraud-detection tools look for

The tools all start the same way: take a sample of a creator’s followers, classify each account, and add up the share that looks real. Where they differ is in what else they fold into the score and how much they tell you about it.

HypeAuditor’s Audience Quality Score

HypeAuditor started in 2017 as an Instagram fake-follower checker, and fraud detection is still what it’s known for. Its Audience Quality Score runs from 1 to 100 and combines 15 metrics in four groups:

• Engagement rate, the share of followers who interact with the creator’s posts.
• Quality audience, the share of real people among the creator’s followers. HypeAuditor separates out suspicious accounts, which it says may be bots or real people using automation services, and mass followers.
• Follower and following growth, looking for abnormal spikes.
• Comment authenticity, the share of recent comments that don’t come from engagement pods.

HypeAuditor’s free Instagram audit shows the score for any account; it calls 60 or more good, and says a score below 40 suggests abnormal follower growth. Its full reports add 35+ vetting metrics. On its homepage HypeAuditor claims it detects 95.5% of all known fraud activity. We haven’t found an independent audit of that figure. HypeAuditor itself says the score isn’t always enough to make a decision, which is the right way to use it.

Modash’s credibility score

Modash’s credibility score estimates the share of a creator’s followers that appear to be real people: 0.8 means about 80% real. Modash says it’s built from profile pictures, bios, follower-to-following ratios, activity patterns and other signals, using network graph analysis. It gives 0.75 as the general benchmark, and notes that benchmarks vary by follower count, so a score can look fine and still trail creators of a similar size. Modash also scores the accounts that like a creator’s posts separately, which is worth checking if you suspect bought likes on an account whose followers look clean.

Its free Instagram fake-follower check, which needs no sign-up, shows the method in miniature: it samples a creator’s followers, sorts each into real people, mass followers, influencers, suspicious accounts and bots, and counts suspicious accounts and bots as fake.

The rest

• Heepsy has an audience authenticity filter and an audience quality score filter from its Plus plan, and free fake-follower checkers for Instagram, TikTok and YouTube. It doesn’t publish how they work.
• Influencity includes a credibility score in its profile analysis, alongside audience quality stats.
• CreatorIQ has audience authenticity tools that flag suspicious activity; the method isn’t published.
• Favikon’s AI Authenticity Score covers LinkedIn and X only.
• influencers.club doesn’t offer a fake-follower score.

Checks you can do yourself

None of these needs a paid tool, and none of them is proof on its own. They’re signals: one is a reason to look closer; three or four together are a reason to walk away. Run them on every creator you’re about to pay, whatever their score says.

1. Look at the growth curve

Organic growth is lumpy but steady: a good post brings a bump, then things settle. Bought followers, as Modash puts it, tend to show up as a sudden spike followed by a flat period. Most databases show follower history, and HypeAuditor’s free report shows gains and losses over the last four weeks. A spike is worth explaining, not damning: a viral post, press coverage or a collaboration all produce one. Ask yourself whether there’s a post that accounts for it. Influship’s profiles carry 30-day growth and engagement rate, so you can flag recent jumps across a whole shortlist before you look closer.

2. Compare engagement with size

Work out the creator’s engagement rate and compare it with creators of a similar size on the same platform. Our engagement rate guide has the formulas and published benchmarks. Both directions are signals. Very low engagement for the size suggests followers who don’t exist or don’t care. Very high engagement on posts that don’t justify it suggests bought likes or a pod. On video, compare views with likes too: thousands of likes on a video with few views doesn’t add up.

3. Read the comments

This is the check tools are worst at and people are best at. Open the last ten posts and read what’s there. Warning signs: short generic comments (“Great post!”, a row of emojis) that could sit under anything; the same handful of accounts commenting on every post within minutes; comments in a language the creator doesn’t post in; and comments that don’t respond to what the post actually says. Real audiences ask questions, argue and tag their friends.

4. Sample the followers

Open the follower list and scroll through the most recent few dozen. Modash lists the traits fake accounts tend to share: no profile picture, a gibberish username, no bio and few or no posts. Accounts following thousands of others are mass followers. A few of these are normal for anyone; a list where they dominate isn’t.

5. Check the audience is where it should be

A creator who posts in English about London restaurants should have an audience mostly in the UK, or at least in English-speaking countries. Databases estimate audience location from a sample of followers; you can get a rough sense from the languages in the comments. A large share of followers in countries that don’t match the creator’s language, content or market can be a sign of bought followers, and even when it isn’t fraud, it’s an audience you can’t sell to.

6. Ask for the insights

Creators can see their own reach, audience location and follower growth. Ask for screenshots from their native insights, or a short screen recording, before you sign. Creators who do a lot of brand work are used to the request. A refusal tells you something too.

What no score can tell you

Fraud scores are good at screening and easy to over-read. Their limits:

• They’re estimates from a sample. No vendor checks every follower of a large account. Two tools can give the same creator different scores, and neither is wrong in its own terms.
• They can’t see intent. A low score says the audience looks low quality, not that the creator paid for it. Giveaways, follow-for-follow and old viral moments all leave junk behind.
• They depend on what’s public. Modash says it can’t estimate audience data when a follower list is private, and hidden like counts limit engagement checks. Private accounts are out of reach entirely.
• Nobody audits them. Vendors publish claims about their own accuracy; we haven’t found an independent test of any of them. Until someone runs one, treat every score as the vendor’s opinion.
• They miss the subtle stuff. Engagement pods made of real creators look like real engagement. Reading comments catches more of them than any score.

Protecting yourself

In the US, the FTC’s Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, in force since 21 October 2024, makes it unlawful to sell, distribute, buy or procure fake indicators of social media influence that the buyer or seller knew or should have known were fake, where they materially misrepresent influence for a commercial purpose. The rule defines those indicators broadly: followers, subscribers, views, plays, likes, saves, shares, reposts and comments all count.

That gives you a clear footing to ask creators about it. In practice: put a warranty in your contract that the creator hasn’t bought followers or engagement, ask for native insights before you pay, and pay at least part of the fee against delivered results where you can. This is practical guidance, not legal advice.

Which tool to use

If fraud is your main worry, HypeAuditor is the specialist: it was built as a fake-follower checker and still leads on vetting depth and platform coverage. Modash’s credibility score is the best documented, and it comes with published prices and a free trial, which is why most teams we’d point at a database start there. Our Modash vs HypeAuditor comparison weighs the two directly, and our influencer database ranking covers the rest:

Questions