Every account of any size has followers it never asked for. Bots follow popular accounts on their own, and a creator can inherit an audience from a clip that traveled somewhere they do not publish. Purchased followers and engagement are real and worth catching, but the goal is not to prove fraud. It is to decide whether this creator is worth a specific amount of money.

No single number is evidence

Look at the growth curve for step changes that no content or press coverage explains. Compare engagement to reach across a stretch of posts rather than one. Examine what the engagement is made of, since likes are cheaper to manufacture than substantive comments. Read the comments themselves for generic, repeated, or out-of-language replies. Check whether audience geography matches the language the creator publishes in and the market you are buying.

Each of these has an innocent explanation on its own. A step change can be a recommendation surge. Off-market followers can come from one video that crossed a border. What matters is whether several signals point the same direction at the same time, and whether the pattern persists across months.

What an audience quality score measures

Third-party authenticity and audience quality scores are model outputs built on sampling and heuristics the vendor chose. They are not a count of fake accounts, and two tools will disagree about the same creator without either one lying. Ask what a score is built from and how it behaves for the platform and creator size in front of you.

Used to rank a longlist, a score saves real time. Used as a pass or fail gate, especially for creators sitting near the threshold, it will reject good partners and approve weak ones with equal confidence.

Decide about the budget, not the person

When the evidence is mixed, adjust the deal rather than the verdict. Reduce the fee, shift weight to a performance component, ask for a smaller first deliverable, or move the creator into a later phase once you have your own data on them.

Record what you saw and when you saw it. A creator who looked wrong six months ago may be clean now, and a documented reason travels through an approval process far better than a shortlist with unexplained gaps in it.