How we verify our data
Where every number on VTuber Profile comes from, how often each figure refreshes, and the verification gate that keeps unconfirmed entries out of search results.
A database is only worth its verification rules. This page documents ours in full — the same rules linked from every talent profile.
Where the numbers come from
- Subscriber counts, channel creation dates, live status — pulled programmatically from official platform APIs (the YouTube Data API and the Holodex API). No scraping of fan wikis, no manual guesses.
- Agency, generation and fan-name attributions — seeded from the public record and then cross-checked automatically: a daily job compares each talent's recorded agency against independent org data and flags or corrects mismatches.
- Schedules — upcoming streams come from official channel listings only. Anniversaries are computed from verified channel creation dates, not typed in by hand.
- Graduations — recorded only on official announcement by the agency or the talent's official channel. Rumors, speculation and "everyone knows" are not sources here.
How often it refreshes
- Live status — every hour. The live board reflects the most recent poll.
- Subscriber counts — daily. Each talent card shows a freshness stamp when data has been synced recently.
- Schedules — daily, with a rolling window so stale placeholder frames and long-past events drop out automatically.
- Roster discovery — weekly. New debuts at tracked agencies are detected from org rosters; independent additions must pass a stricter match check before entering the database.
The verification gate
Every talent entry is either verified or unverified. Verified means the official channel has been positively identified and confirmed against thresholds (channel identity match, plausibility checks on the data). Unverified entries — new additions the pipeline hasn't confirmed yet, or ambiguous matches — remain visible internally but are excluded from search engine indexes until they pass. We would rather show fewer pages than wrong ones.
When automated cross-checks disagree with our records, the entry is corrected if the evidence is unambiguous, or flagged for review if it isn't. Corrections are logged reversibly, so a bad automated decision can be rolled back.
What we deliberately do not publish
- No hosted artwork. Character art, stream captures and thumbnails are copyrighted; the only visual identity on this site comes from official channel embeds. Details on the copyright page.
- No real-identity information. Names, faces and past identities of performers are out of scope, permanently — see the graduation guide for the cultural context.
- No estimated figures. If an API doesn't return it, the field stays empty. A dash on a profile means "not verified", never "we rounded".
- No speculation on departures. Status changes follow official announcements exclusively.
When we get something wrong
Automated pipelines make mistakes; the fix path is public. Any page can be challenged through the correction request form — corrections with a source are applied against the same verification rules above. The source policy describes precedence when sources conflict.