OnlyDetect's 30 September 2026 audit found 3,625 tag assignments across 447 of 461 eligible profiles, but no assignment with a recorded verification date. Tags were common in the directory; recorded verification was absent. Those are different measurements.
The finding means these tags should be treated as discovery leads, not evidence that a creator has personally confirmed a category or that every tagged result matches an offer.
Coverage and verification answer different questions
Coverage asks whether a record has at least one tag. In this sample, 447 did and 14 did not. Verification asks whether the dataset preserves evidence of a check. The verified_at field was empty for all 3,625 assignments at the observation time.
The aggregate dataset and SQL make those definitions explicit. An empty verification timestamp does not prove a tag false; it means this particular audit cannot claim a recorded verification.
Scroll across the table on smaller screens.
| Measure | Observed value |
|---|---|
| Eligible profiles | 461 |
| Profiles with one or more tags | 447 |
| Total assignments | 3,625 |
| Assignments with verification timestamps | 0 |
| Median number of tags per eligible profile | 3 |
| Largest recorded tag count on one profile | 118 |
Why more tags do not establish a better match
Tagging practices can vary between sources. A record with many tags can appear in more filtered results without being a more suitable profile. The observed maximum of 118 should therefore be understood as a data characteristic, not a recommendation.
A useful matching system needs to consider provenance, conflicting evidence, and the intended meaning of a category. Simply counting tags rewards broad labeling. It does not validate the underlying claims.
What fans can do with a tagged result
Use the result as a starting point. Read the relevant public source and check whether the creator actually describes the topic or offer that matters to you. If the evidence is absent, keep the result uncertain rather than treating the tag as a promise.
The guide to official profile links explains the separate task of confirming the account connection. A correct account link still does not prove that every third-party tag is accurate.
What this means for AI recommendations
An AI system does not turn an unverified tag into a verified fact. Recommendations should preserve the underlying evidence level and be able to say that the available fields are insufficient.
OnlyDetect's AI discovery guide describes the limits of matching from public profile data. This audit measures a specific data gap; it does not report the accuracy of all AI recommendations or the performance of competing services.
How the finding could improve
A future review could sample assignments, compare them with current creator-controlled descriptions, record the result and date, and measure agreement. That would be a new study with its own denominator. Until such evidence exists, reporting coverage separately from verification is the accurate conclusion.