HomeFact-Check › Fact-check a claim or video

How to fact-check a claim or YouTube video with AI

Verify any claim, article or video and get a 1–10 truth score backed by real, linkable sources — cross-checked by two independent AIs, with a reasoning trail you can audit.

Manually verifying a single claim can take many minutes of searching and cross-referencing. AI fact-checking speeds this up, but a bare verdict from one model is not trustworthy on its own. The approach that holds up is evidence-first: two independent AIs check each claim against live web sources, you see those sources, and the result carries a short note on how strong the evidence actually is.

Step by step

  1. Select what to verifyHighlight a claim on any page, or open the article or YouTube video.
  2. Start the checkRight-click and choose fact-check, or use the toolbar button. For videos, the tool reads the captions — no download needed.
  3. Read the sourced scoreA 1–10 truth score with real sources, an inline evidence-quality note, and a "why this score" reasoning trail.
  4. Re-check or go deeperRe-run against a different source set, or use Full Spectrum for steelmanned counter-views, source diversity and open questions.
🧭 Why two AIs: a single model can be confidently wrong. Cross-checking each claim with two independent AIs (Google Gemini + xAI Grok) surfaces disagreement instead of hiding it — when they diverge, you see it, which is exactly when a human should look closer.

Frequently asked questions

How can I fact-check a YouTube video?
The tool reads the video's captions and verifies the claims in them, returning a 1–10 truth score with sources. You just point it at the video — no downloading or re-uploading.
How accurate is AI fact-checking?
It's a research aid, not a final authority. Using two independent AIs that cross-check against live sources — and showing those sources and the reasoning — is far more trustworthy than a single unexplained verdict. Always be able to inspect the sources.
What does the 1–10 truth score mean?
It summarizes how well evidence supports the claim: low = contradicted/unsupported, high = well-corroborated. An inline note states the key caveat, e.g. single source vs. multiple independent sources.
Is my data private?
Only the text you explicitly submit is checked, and personal data (emails, phone numbers, card numbers) is stripped locally first. Nothing runs in the background.

🛡️ Verify with wyper.io Fact-Check

A 1–10 truth score with real sources, two-AI cross-check, and an auditable trail — on any page, selection or video.

See how it works →