Influencer marketing buys audiences by the number, and the number lies constantly: inflated by dead subscribers, giveaway tourists and view-farming formats that reach many and move nobody. Meanwhile the metric that predicts campaign performance — does this creator's audience actually trust them on this topic — appears on no dashboard.

Trust shows in the content: creators who explain rather than perform, who've covered the category with substance over time, whose recommendations carry first-hand experience. It's readable — from what they say, not from their subscriber count. Which means vetting can rank candidates by the thing that converts, if it reads the content instead of the media kit.

deep_search builds a candidate list on substance: creators covering your category, with quality verdicts doing the credibility triage — original (first-hand experience, not rehash), factual (claims backed, not vibed), high_density (substance per minute). The niche-reach dial surfaces the mid-size creators where trust-per-dollar peaks.

library_import_youtube_channel plus deep_context deep-vet the finalists: import a candidate's channel and interrogate the fit — how do they talk about products in your category? how did past sponsorships integrate? does their expertise claim hold across the archive? — cited to the videos. The shortlist that survives ranks by demonstrated credibility on your topic, which is the number the follower count was always pretending to be.