Methodology · v1
How the score works
No black box. Here is precisely what goes into a So Influential authenticity score — and what we do when the data is thin.
The four categories
Audience authenticity
Share of followers with no photo, no posts, bot-pattern usernames, or off-audience language; plus follow:follower imbalance on the account itself.
Engagement quality
Engagement rate vs the median for the same platform and size band — both far-below (dead/bought audience) and far-above (pods, bought likes) lower the score. Like:comment sanity is checked.
Growth pattern
Follower history scanned for sudden spikes — step jumps that make up an outsized share of total growth read as bought followers.
Content & brand safety
Account age, sponsored:organic ratio, and brand-safety keyword scan on recent captions.
Honest confidence
When we can only see part of the picture — a small follower sample, no growth history — we lower the confidence of that category and let it pull less weight, rather than inventing precision. That's why a thinly-sourced account shows a confidence note instead of a falsely exact number. It's the single biggest thing the incumbents get wrong.
Bands
- 80–100 Authentic — healthy audience, normal engagement.
- 60–79 Mostly real — largely genuine, a few signals to check.
- 40–59 Mixed — vet manually before paying.
- Below 40 High risk — strong signs of fake followers or bought engagement.
Versioned & reproducible
Every score is stamped with a methodology version (v1). When we improve the signals we bump the version, so a cached score always tells you exactly which logic produced it.
See it on a real account
Run a free check — the result page shows every category and its confidence.
Check a creator →