I had a sobering moment last month. We were working with a creator who checked all the boxes in our vetting system: engagement rates looked solid, audience demographics matched, language skills were bilingual. But about two weeks into the campaign, we started noticing something weird. The comments were growing, but they felt… off.
Turns out, this creator had been slowly purchasing engagement over the past few months—nothing dramatic enough to trigger obvious red flags, but enough to inflate the authenticity score. And here’s the kicker: because the creator was bilingual, they had accounts in both English and Russian, and the fraud patterns weren’t immediately obvious when you looked at each market separately. It only became clear when we looked at the cross-market behavior.
That experience made me realize something: when you’re vetting creators across multiple markets and languages, traditional fraud signals might not be enough. A creator could look legitimate in one market while their behavior in another market tells a completely different story.
I’ve been thinking about what the blind spots actually are in discovery tools:
First, engagement velocity. Real growth happens gradually. When you see sudden spikes in followers or engagement across one or both markets, that’s worth investigating further, even if the absolute numbers look reasonable.
Second, audience composition consistency. If a creator’s English-speaking audience has totally different demographics than their Russian-speaking audience, that can be a signal. Sometimes it’s legit (they actually serve different communities), but it’s worth digging into.
Third, comment language matching. This one caught our fraud case. Real followers comment in their native language naturally. If you see comments in a language that doesn’t match the creator’s typical audience, that’s a red flag we weren’t explicitly checking for.
Fourth, content adaptation authenticity. A bilingual creator should adapt content for different markets—different references, different tone, different pacing. If everything looks like a direct translation, that’s worth questioning. Same goes if the tone is completely different between languages (suggesting maybe different people are managing accounts).
The hard part is that none of these signals alone means fraud. It’s the pattern that matters. But most tools check individual signals in isolation, not in combination.
Have you run into situations where a creator looked legitimate in your primary market but sketchy in a secondary market? How did you actually catch it?