Video is the hardest medium to vet

Text is searchable. Video is not. A single interview, panel appearance or livestream can run two hours, and there is no way to keyword-search speech you have not already transcribed. Anyone screening a public figure, a prospective hire or a partner organisation faces the same problem: the risk is spoken aloud somewhere in twenty hours of footage, and the only traditional method of finding it is watching all of it.

ArgusX (also written Argus X) closes that gap. You submit the video URLs you want reviewed, ArgusX transcribes each one in full, and then runs the transcript through the same AI classification pipeline used for social posts. Hours of viewing compress into a scan that typically finishes in one to fifteen minutes, and you are emailed when the dossier is ready.

What the YouTube scan actually does

You control the scope: ArgusX analyses the videos you paste in rather than crawling a channel on its own, so you always know exactly what is being covered and what it will cost. Each video is transcribed end to end, so a remark made ninety minutes in is treated the same as one made in the opening seconds.

A single scan covers up to twenty hours of total runtime. Video length is checked before the scan starts, so you see how much of your allowance a batch will use before you commit to it.

The same nine risk classes, plus your own

ArgusX screens transcripts against nine reputational risk categories — racism, violent rhetoric, antisemitism, Islamophobia, transphobia, homophobia, sexism, hate speech and extremism — and surfaces every quote that trips one, with the flag attached. You can also add a custom query for a topic, person or position specific to your research, at no extra cost, and run it alongside the standard categories or on its own.

Because the transcript is screened rather than skimmed, YouTube risk screening catches language that a channel's titles, descriptions and thumbnails would never reveal.

Timestamped, verifiable findings

Every flagged quote comes with a plain-English rationale explaining why it was surfaced and a link to the exact moment in the source video. That is what makes the output usable outside your own notes: a client, an editor or a hiring panel can click through and hear the quote in context rather than taking the classification on trust. Each scan also produces a video summary and an executive overview, and the whole dossier exports as a PDF.

Transcription and analysis run in stages. If a stage fails — a video that will not yield audio, for example — a rescan resumes from where it stopped instead of repeating work, and the results you already have stay on screen while it runs.

Where video analysis fits in a wider review

In practice, video rarely stands alone. A full picture of a subject usually pairs their spoken record with their written one, which is why ArgusX also lets you scan an X (Twitter) account's full history and covers Bluesky and Rumble with the same classifier. Teams running structured screening programmes should start with our social media vetting overview, and political teams with opposition research workflows.

One scan is one credit — for a set of videos as much as for an account. Subscriptions include monthly credits and shared seats, and credits can be bought individually; see pricing. Creating an ArgusX account requires no card, so you can look at the scan configuration first.