ChamsPilot · Methodology
We publish one measured figure for creator verification: in a blind test on 640 simulated creator accounts, 95% of High Risk flags were correct. This page explains how that test was run, what it shows and what it does not, and where our audits read official data or make an estimate.
ChamsPilot is an approved API partner with Google, TikTok, Meta and X. Where a creator's real figures are available through those official APIs, we read them directly instead of inferring them.
That path is open more often than people expect. A creator listed in a platform's own creator marketplace, using TikTok One, or who has run collaborations tagged as paid partnerships, has data the platform already confirms. For those accounts the audience numbers are the platform's record, not our estimate.
Platform data tells you what an audience is. It does not tell you whether the engagement is real. Every audit also runs our own analysis: comment authenticity, follower composition, engagement patterns and timing, cross-checked by several algorithms rather than one score.
The published figure measures this analysis: the verdict engine behind the dashboard verification and the AI Deep Audit. It does not describe the free basic check on our fake follower checker pages, which uses a lighter heuristic.
We built 640 simulated creator accounts, 160 each on TikTok, Instagram, YouTube and X. Half grow organically, including the cases that trip up fake follower checkers: giveaways, fandoms, crossovers, comment-keyword calls to action, low engagement and non-English audiences. The other half inflate their numbers in several different ways: bought followers, likes, views or comments, engagement pods and follow-for-follow.
The verdict engine was frozen before the test, and the sealed set was run once. The results below come from that single run.
In the blind test, the engine flagged 166 of the 640 simulated accounts as High Risk, and 157 of them were inflated: 95% of High Risk flags were correct. 9 of the 320 organic accounts were flagged High Risk by mistake, a false positive rate of 2.8%.
It is built not to accuse real creators, so it flags less than it could. Of the 320 inflated accounts, 157 (49%) were flagged High Risk, 47 more were marked Needs Review for a closer look, and 116 passed. When the same blind test on the 640 simulated accounts added two optional lookups, a sample of recent followers and a daily follower history, 66% were flagged High Risk and 95% of flags were still correct.
What this test does not show: the accounts are simulated, not real creators, and they were generated from the same picture of organic accounts that the engine's norms come from, so results on real accounts may be lower. A benchmark on real accounts, labelled independently before ChamsPilot analyses them, is being prepared and will be published on this page.
For the remaining accounts no official data is available. We still audit them, and the comment and engagement analysis works the same way. But some attributes cannot be read, only inferred, and the clearest example is audience geography.
When a platform does not supply follower and viewer countries, we estimate them: correlating comment timing against timezone activity, alongside other signals we have built and keep refining. It is an estimate, and we label it as one.
No third party can match a platform's own data about its own users. Any service claiming otherwise is guessing. Where the platform gives us the real figure, we show it. Where it does not, we show our best estimate and say so.
Creators use it to understand their own numbers and improve them. Seeing which part of an audience is genuinely engaged is the first step to growing the part that is.
Brands, agencies, CEOs and professionals use it to judge what a creator is actually worth before money moves: real reach, visible data, and a defensible reason for the rate.
Read in another language: 简体中文 繁體中文 Français Español العربية