Press Kit

Data source. Our scanner queries Odysee's public claim_search API (the same endpoint powering odysee.com's own search) across ~170 keywords spanning 14 violation categories: terrorism, hate speech, extremism, harassment, violence, CSAM, drug trafficking, weapons, fraud, cybercrime, human trafficking, animal cruelty, self-harm, and conspiracy.

Classification. Items are categorised on title content only, using word-boundary regex matches against a published keyword map. Items retrieved by a violating query but whose titles do not independently contain a trigger phrase are labelled "Under Review" rather than over-classified. Our reported category counts are a conservative floor, not a ceiling.

Blocklist. To suppress obvious false positives (science channels, gaming, cooking, etc.) we maintain a published blocklist of channel handles and title patterns. Source is visible on GitHub.

Cadence. The scanner runs twice daily via GitHub Actions, with polite request delays that stay well under Odysee's rate limits. Every scan result is committed to the public repository and triggers a Netlify redeploy.

Scope limits. The scanner only reads public metadata (titles, channel handles, effective LBC stake). It does not view, download, or mirror any video content. It does not access private channels or bypass any access controls.

Reproducibility. Scanner source, keyword list, blocklist, and daily results are public at github.com/bryanflowers/odyseeexp. Any journalist or researcher can re-run the scanner end-to-end.