Reach is only worth what the people behind it are worth. Audience Credibility answers the question every buyer has before spending: how much of this audience is real people who can actually see and buy things? Every account in the engaged audience is classified into one of six types, and the result rolls up into one score from 0 to 100 that reads directly as a percentage: a score of 84 means 84% of the analyzed sample looks credible.
Start From the Engaged Audience
For each creator, we analyze a sample of their engaged audience. Credibility is exactly where this matters most, because low-quality accounts are built to inflate engagement. Audience estimates are extrapolated from public engagement signals, and every audience statistic is an aggregate.
The Audience We Analyze
Every audience data point we publish is built on the people who actively interact with a creator's content, because they're the ones who see and respond to campaigns. The full reasoning behind this approach is on the Audience Analytics page.
Six Kinds of Accounts
Every account in the sample is classified into exactly one of six types, based on public signals: whether it has a profile picture and posts, how its follower and following counts relate, how long it's existed, and whether it shows real activity.
The Account Types
The classification judges individual accounts in the audience, not the creator. A creator with a high suspicious share isn't accused of anything: low-quality accounts attach themselves to popular content everywhere. The score simply tells you how much of the reach is real.
From Types to a Score
The six types roll up into shares of the analyzed sample, and the score is the credible part: real people plus influencers. That's why the score reads as a percentage. A class label from Very low to Very high puts the score in context, and the result only ships above the minimum sample floor: at least 1,000 engagers, and up to 10,000 for creators with larger audiences.
The Audience Type Split and the Score
Credibility patterns differ by account size: a million-follower creator naturally attracts more low-quality accounts than a 20K one. The class label (Very low, Low, Normal, High, Very high) is calibrated against creators in the same follower range, so "High" always means high compared to the creator's actual peers, not against an unfair global bar.
Illustrative Example
Two creators with similar reach on paper, and very different amounts of it that's real.
Two Creators, Side by Side
84% of the analyzed audience is real people and influencers. Very little of the paid reach would be wasted on empty accounts.
Just over half the audience is credible: mass followers and inactive accounts eat 34% of the reach. The follower count overstates what a campaign would actually touch.
If these two creators charged the same rate, NikolaTravels would deliver roughly half again as much real reach per dollar. That difference is invisible in follower counts and engagement rates, which is exactly why this score exists.
Output Format
The exact fields this data point ships with, shown with example values, and how to use them in search.
The Actual Fields
The six type shares sum to 100% of the analyzed sample. The score equals real people plus influencers.
Filter by score for a hard floor, for example "credibility above 75" to keep wasted reach out of a paid campaign. Or filter by class ("High" and above) to apply a bar that's fair across creator sizes. The type split is there when you want to know why a score is what it is.
Validation 
Audience Credibility is validated at both levels: the account classification underneath, and the aggregated scores built from it. Both go through the same two-track process used for every data point.
Getting individual classifications right is not enough. We also validate the final scores, comparing published credibility against independently built reference data for the same creators. This catches errors that only show up in aggregate, and it is why the field ships with a sample floor instead of shipping every creator with any data at all.