Evidence notes
Clear, practical reading for public social clips: what to check first, what a detector can and cannot tell you, and when the claim needs human verification.
Latest field guide
Video provenance guideTrace a suspicious video's earlier source with keyframes, visual search, quoted dialogue, and a source ledger—without treating a match as proof.
Read the checklist
Video provenance guideCheck what video Content Credentials actually declare, interpret missing or invalid provenance, and combine C2PA information with independent evidence.
Read the checklist
Detector result guideRead likelihood, confidence, checked signals, source quality, and provenance together—never treat one detector percentage as proof.
Read the checklist
X / Twitter verification guideCheck the post, source, provenance, continuous motion, and independent context before treating an X video—or a detector score—as proof.
Read the checklist
Deepfake verification guideVerify a suspicious clip with source context, provenance, cautious screening, and corroboration—not a detector score alone.
Read the checklist
Evidence-quality guideCompression changes the evidence available for review. Use source quality, context, and a cautious screening report instead of a single visual tell.
Read the checklist
YouTube verification guideUse a public YouTube link as a first-pass screening input, then read confidence, limitations, source context, and next steps together.
Read the checklist
Detector troubleshooting guideA limited result means the available link or media did not support a fuller screening assessment—not that the video is real or fake.
Read the checklist
Instagram verification guideUse source context, disclosures, the earliest version, careful frame review, and cautious detector evidence before sharing a suspicious Reel.
Read the checklist
TikTok verification guideSeven practical checks for TikTok labels, source history, repost damage, frame patterns, audio, context, and cautious detector results.
Read the checklist