I audited 43,000 Indian Instagram influencers. 27% have fake followers.

Six months ago, I had a hypothesis I really didn't want to be right about: a meaningful chunk of India's "top creators" have followers no brand will ever convert. I built the tooling, ran the data, and now I have an answer. It's worse than I thought — but the patterns are predictable enough that you can avoid almost all of it in about ninety seconds per profile.

This is what we found. The numbers, the methodology, the names of the patterns (not the creators — I'll explain why), and a checklist you can use today.

What we did

Between November 2025 and April 2026 we indexed 43,231 Indian creator profiles across Instagram, YouTube, and TikTok. The Instagram set — the focus of this piece — was 36,000+ accounts pulled through a combination of public profile data, the Meta Graph API via connected business accounts, and engagement signals from the last twelve posts on each profile.

For every account we calculated:

• Follower-to-engagement ratio
• Comment quality (length distribution, emoji density, vocabulary diversity)
• Day-over-day follower growth velocity
• Engagement decay relative to reach
• Cross-platform presence (Twitter, YouTube, TikTok)

We then ran the result through a gradient-boosted classifier trained on 487 manually-audited "ground truth" accounts. Our internal validation came in at 94% agreement with manual audit — meaning when our system flagged an account as "15%+ fake," a human auditor agreed about 19 times out of 20. Not perfect. But useful enough that a brand spending five lakhs on a campaign should care.

The fraud rate is shaped like a U

Here's the table I wish I'd had two years ago. Fraud rate by tier:

The micro tier — the one every D2C brand is told to focus on for "authentic engagement" — is the most likely to be lying to you. That's not because micro-creators are uniquely shady. It's because the gap between 9,000 followers and 50,000 followers is the gap between "I post for fun" and "I get pitched by brands." A few thousand bought followers is the cheapest possible bridge. Eight hundred rupees on a Telegram seller solves it.

The five signals that catch 90% of fraud

You don't need an AI model. You need ninety seconds and a pen. Here's what to look at:

1. Engagement-to-follower ratio (the obvious one)

Take an average of the last 12 posts. Add likes plus comments. Divide by follower count. Multiply by 100. That's the engagement rate.

Real creator: 3-7% (micro), 1.5-4% (mid), 0.8-2.5% (macro), 0.4-1.5% (mega).
Fake-padded creator: usually under 1% even at 50K followers.

Why it works:bots follow but don't engage. The more you buy, the more your ratio collapses.

2. Comment-to-like ratio

Real audiences comment about 1.5-3% as often as they like. So a post with 10,000 likes should have 150-300 comments. If you see 10,000 likes and 20 comments, something is off — bots love but don't talk.

3. Comment quality (the one nobody checks)

Open the comment section on three recent posts. Read the first 30 comments out loud. If more than 60% are:

• Emoji-only (🔥🔥, ❤️❤️❤️)
• Under five characters ("nice", "wow", "cute")
• Generic praise ("Beautiful pic", "Looking great")
• From accounts with under 200 followers themselves

...you're looking at engagement pods or comment-for-comment groups, not real fans. The median real comment is 47 characters and references something specific in the post.

4. Growth velocity

Tools like Social Blade show daily follower history. A real account grows in a wavy line — viral spikes, slow decay, plateau, repeat. A bought account looks like a staircase. Sudden +5,000 followers in a day with no corresponding viral content? That's a purchase.

The most damning version: follower count goes up, average engagement on new posts goes flat or down. Real growth pulls engagement up with it. Bot growth doesn't.

5. Cross-platform shadow

This is the single sharpest signal we found. If a creator has 500K Instagram followers but zero Twitter, zero YouTube, and zero TikTok presence — the probability their Instagram is inflated is 73%.

Real creators diversify. They protect themselves against platform risk. They cross-post. A creator who built 500K real followers on Instagram has either tried other platforms or has a website or is at least mentioned in news articles. A pure-Instagram account with no traceable identity outside the app is a giant red flag.

Why the agencies don't tell you

I want to be careful here because most influencer agencies in India are doing fine, ethical work. But there's a structural problem worth naming.

Agencies make a percentage of the campaign spend. The bigger the spend, the bigger the cut. A brand that says "we want to spend ₹50 lakh on a creator with 1M followers" is more valuable to an agency than a brand that says "we want to spend ₹50 lakh on five creators with 100K real, engaged followers each." The first is easier to fulfil and bills the same. The second requires more vetting, more contracts, more reporting.

So the conversation drifts toward the bigger names. And the bigger names are sometimes — not always, but sometimes — exactly the ones with the inflated metrics. The agency isn't lying. They just aren't looking very hard. And nobody in the room has a strong incentive to ask "are these followers real?" until the campaign disappoints.

By then the money is gone.

The mega-celebrity problem (different fraud, same waste)

Here's a thing I didn't expect to find. Bollywood stars and Indian cricketers — the very top of the pyramid — usually have the cleanest follower bases in the country. Their fake-follower rates are 4-12%. Way below the median.

But their engagement rates are terrible. Often under 1%. Sometimes under 0.5%.

Why? Because they have 80 million followers, but only a small fraction of those followers are actively interested in this post about thisbrand. The follower count is real. The reach isn't.

What this means in rupees: a Bollywood star at ₹3 crore per post might reach 4 million people, generating an effective CPM of about ₹750. A 200K-follower beauty creator at ₹3 lakh per post reaches 70K people in her actual buying audience for an effective CPM of ₹43. Even if you account for higher conversion on the celebrity (more trust, more aspiration), the math doesn't close.

Mega-celebrities are not a fraud problem. They're a unit-economics problem.

Why I'm not naming creators

I get this question every time I show this data privately. "Just publish the names. People deserve to know."

Two reasons I won't.

One— the methodology has a 6% error rate. That means if I publish a list of 1,000 names, around 60 of them are wrongly flagged. Those 60 careers don't deserve the hit. Not a tradeoff I'm willing to make.

Two— the goal isn't to take down creators. The goal is to give brands a tool. Every creator on Dexfluence has their authenticity score visible to the brand looking at them. The score updates weekly. If a creator cleans up — stops buying followers, focuses on real engagement — their score recovers. That's the right outcome. A name-and-shame list locks people into their worst week.

What I'm happy to share publicly is the aggregate picture and the methodology. Look at the patterns, then verify your specific shortlist before you sign anything.

What this means for your next campaign

Three things I'd do differently if I were running brand campaigns again with this data:

Triple your shortlist.If you needed five creators, longlist fifteen and audit aggressively. The 27% fraud rate means roughly four of those fifteen will fail the basic checks. You'd rather find that out before signing.

Pay for the right thing. Stop paying based on follower count. Start paying based on average engaged audience size — likes plus comments per post times some multiplier. A creator with 100K followers and 6% engagement is delivering more real reach than a creator with 1M and 0.5%, and they should price accordingly.

Keep your micros honest.The micro tier is where you find the best ROI in India — but it's also where the highest fraud rate sits. So your due diligence has to be tightest exactly where you're most excited. Run all five signals on every single micro before you sign.

The final spreadsheet test

If you take one thing away, take this. Put your shortlist in a spreadsheet with these columns:

Three out of five pass = green light, proceed with diligence.
Two of five = yellow, ask harder questions before signing.
One of five = red, walk away.

That's the framework. It's not magic — it's twenty minutes of work per creator. Most agencies aren't doing it. Doing it puts you ahead of 80% of brands in India.

One more thing

We baked all of this into Dexfluence — the platform I'm running. Every Indian creator in our database has an AI authenticity score from 0-100 visible to brands. We update it weekly using the same five-signal model from this piece. If you want to outsource the spreadsheet, that's what we're here for.

But honestly? Whether you use us or not — please run the checks. Your campaigns will be better. The Indian creator economy will be cleaner. And the next ₹460 crore won't end up in bot farms.

That's a worthwhile trade.