The score answers one question: how many real people does this creator actually reach? It’s a number from 0 to 100, built in three moves: start with audience size, subtract what looks fake, and weigh how strongly the audience shows up.
We track each creator’s followers on X and subscribers on YouTube, refreshed continuously from the platforms. A creator’s biggest platform counts in full and each additional platform adds 25% of its audience, because a lot of the same people follow a creator in both places and counting them twice would reward simply being everywhere. If a creator hasn’t posted on a platform in 90 days that platform doesn’t count at all.
This is the part nobody else does. Every month, for each major creator, we pull real accounts from their audience and look at each one: how old the account is, whether it has ever posted, whether anyone follows it back, whether it still has the default blank avatar.
An account only counts as fake if it’s unmistakable — for example an account that follows 2,000 people, has almost no followers of its own, and never posts; or one that has literally nothing: no photo, no posts, no followers.
The share of the sample that’s clearly fake becomes a straight discount: if 20% of the sampled audience is fake, the creator’s audience number is cut by 20%. If 50% is fake, it’s cut in half. (The discount is capped at 80%, and it doesn’t apply at all until we’ve accumulated a large enough audience base for that creator to limit the margin of error.)
A million followers who never watch, like, or reply aren’t a million people you reach. So we measure engagement per view across everything a creator posted in the last 90 days, with replies and comments counting the most (they’re hardest to fake), then shares and bookmarks, then likes. Recent posts count more than older ones, and we use the median post rather than the average. While viral posts measure that creator’s reach, it doesn’t help us answer the question how many accounts included in the reach are real people.
Each creator is then compared to creators of similar size on the same platform, so a million-subscriber channel is judged against other million-subscriber channels, not against a 20,000-follower account. The least-engaged creator for their size has their audience number trimmed by up to 30%; the most-engaged loses nothing. This method accounts for the fact that larger accounts receive far less engagement for benign reasons (far more people want to “lurk” a large superstar than a small creator). Engagement is another way to see if an audience is real, and we want to confirm our hand-checked numbers from step 2 match with engagement data. If there’s a large discrepancy in either direction it’s a reliable sign the apparent audience doesn’t match that creator’s real audience.
A creator’s followers jumping in a week without a corresponding move in their views and engagement over the coming weeks marks the signature of bought followers, since real growth brings real eyeballs. We have automatic checks looking for this and docking new followers under this trend.
After the checks, each creator has a “credited audience” which equals audience size × share that’s real × engagement rate (as mentioned above, we don’t dock more than 30 percent for poor engagement). Audiences range from thousands to tens of millions, so we place that number on a fixed scale where roughly the 99th-percentile creator sits at 100. The scale doesn’t move week to week — a 78 next month means the same thing it means today. Rankings refresh every Monday!
Honestly: this is version one, and we treat it that way. The score runs on everything the platforms let us collect — the numbers on public record, plus the real accounts we sample and check every month. That evidence grows monthly, and the methodology keeps improving as it teaches us. Two commitments while it matures: when the evidence about an account is ambiguous, we count it as real, so if the score errs, it errs in the creator’s favor. And nothing here is a black box — the full method is above, and every discounted score traces back to real accounts we actually looked at. Given all the information we’re able to collect, we are continuing to build the methodology to make this the most accurate way to verify the true audience of a political creator. If you see something that looks wrong, tell us — checking is the whole point.