Notes on how verifAI works under the hood, what we learn from running thousands of fact-checks, and practical guidance on media literacy for short-form video.
"Unverified" is not a euphemism for false, and it is not a cop-out. It is the honest state of a claim when the evidence needed to confirm or refute it is absent or not yet in.
Read article →A specific vote count feels authoritative enough to share without checking. This is the story of one reel, one tally, and what the actual congressional record showed.
Read article →When ideology wears the clothes of fact, standard fact-checking tools hit a wall. Here is how verifAI handled a sweeping economics claim — and where it honestly had to say "unverified".
Read article →Two very different problems, one easy mistake: old footage recaptioned as breaking news, and fully synthetic video. Here is how to tell them apart — and where both checks hit a wall.
Read article →A reel claimed Madrid never applied Spain's Housing Law — verifAI found two claims true, one false, and five unverified. Here's what the evidence actually shows.
Read article →Invoking "a study" borrows authority without accountability. Here is what a real citation needs, how bad actors exploit the gap, and why verifAI returns "unverified" instead of guessing.
Read article →A practical, step-by-step guide to verifying a video claim by hand — the same logic verifAI runs automatically, laid out so you can do it yourself right now.
Read article →Before you share that reel, check for these five patterns — they show up in misleading short-form video far more reliably than any single claim.
Read article →A walkthrough of the actual pipeline that turns a pasted reel URL into a verdict — from yt-dlp download to Whisper transcription, visual-text extraction with Llama-4 Scout, claim splitting, and the cross-check against Google Fact Check Tools and live web sources.
Read article →Google's Fact Check Tools API is excellent — and it covers only a small fraction of the claims that actually show up in Instagram reels and TikToks. Here's the data we see from the verifAI pipeline and why we treat it as a starting point, not the answer.
Read article →Not every sentence in a reel is something you can fact-check. A practical taxonomy of verifiable vs. unverifiable statements — useful both for everyday readers and for anyone building a fact-checking pipeline.
Read article →Posting a fact-check in the comments of a viral reel feels productive. Most of the time it does nothing. A few small choices about timing, tone, and what you link to make a measurable difference.
Read article →