We screened 95,577 creator accounts: Threads has the highest share of suspected inflated followers at 15.5%

The most expensive mistake in creator marketing is not picking the wrong person — it is paying for bought followers. We ran follower-structure and engagement-ratio anomaly detection across 95,577 creator accounts in our index. Results by platform below.

The headline: the gap between platforms is large — 15.5% on Threads versus under 1% at the bottom. That does not make a platform "bad"; it means the vetting cost you should budget differs enormously by platform.

Threads broken down by follower tier

How to use this number

Do not use it to pick a platform. A high rate does not mean the platform is bad — it mostly reflects how mature the follower-selling market is there, and what kind of accounts our collection tends to surface on that platform.

Do use it to size your vetting effort. On Threads, treat manual verification as a necessary cost: check whether follower growth came in vertical steps, whether the comments are the same handful of accounts posting templated praise, whether the view-to-follower ratio sits persistently off-band. On a platform under 1%, that same effort is largely wasted time.

The most reliable signal only arrives after you collaborate. Any static metric can be gamed on purpose; only a small tracked pilot tells you whether clicks became real users.

Methodology

The sample is 95,577 creator accounts indexed and validated inside Koinon Link. It is not every creator on these platforms, and it is not a random sample — it comes from our keyword-driven collection, which skews toward AI software, e-commerce and gaming.

"Suspected inflated" is our own anomaly detection, not a platform-official flag: it looks at whether follower growth arrived in vertical steps, whether the engagement-to-follower ratio sits far outside the normal band for that platform and size tier, and how templated the engagement looks. It means "worth a manual check", not "this account is definitely cheating".

All figures are recomputed daily; this page shows the most recent run.