Source Catcher
The order professional verification teams work in — and how to do each step yourself, for nothing.
Most people start by staring at the footage looking for something that seems fake. That is the last thing a verification team does, and usually the least useful. Here is the order they actually work in, and why.
The most common form of video misinformation is not a deepfake. It is real footage with a false caption — genuine video from another year, another country, or a feature film, presented as something happening now. Deepfakes get the attention; recycled footage does the damage.
Which means the useful question is rarely "is this AI?". It is "where is this actually from?"
Almost every clip circulating is a piece of something bigger — a film, an episode, a match, a broadcast, a press conference. Establish that and most arguments end immediately.
Look for what is in the frame rather than what the caption says: a logo in the corner, a chyron, a scoreboard, a recognisable face, a set, a title card. A watermark from a different platform is particularly telling — a TikTok handle on a video posted to X means it was lifted from somewhere else.
This catches more misinformation than anything else on this page. Old footage relabelled as breaking news is the single commonest trick.
A clip claiming to be from a conflict that started last month, showing trees in full leaf during that region's winter, has answered the question by itself.
Reverse image search does not work on video, so take stills and search those. This is the technique behind most published debunks.
Search the claim, in the words a fact-checker would use, on Google Fact Check Explorer or the site of a member of the International Fact-Checking Network — Full Fact in the UK, and equivalents elsewhere.
A study of a thousand false claims found 84% returned nothing from the main public fact-check database. Silence is not a clean bill of health. It usually means nobody has got to it yet.
The same study found that the same claim, worded two different ways, produced different results 81% of the time. If the first search finds nothing, try rephrasing before concluding anything.
Last, because it is the least reliable step and the most likely to mislead you in both directions.
What actually helps: hands and fingers, teeth, the way hair meets a background, text on signs and clothing, reflections, and whether physics behaves — water, cloth, crowds. What does not help: a general feeling that something looks "off". Compression makes real footage look wrong all the time.
Calling real footage synthetic is as damaging as missing a fake, and it happens far more often than people expect. Content credentials would settle it, but every major platform strips that metadata on upload, so the evidence is usually gone before you see the clip. If you are not certain, say so.
Working from step one downward, most clips are settled before you reach step five. Working the other way — staring at the footage first — is how people end up confidently wrong about a real video, and how genuine misinformation slips through because the fake looked convincing.
Do all five automatically — free, no sign-upSource Catcher runs these steps on a link you paste: names what the clip was cut from, corroborates it against independent reference works, finds when it first appeared, reads the picture for its age, transcribes what is said, and gives you the frames ready to search. It is free and there is no account.
It also tells you when it does not know, which is the part most tools skip. What it gets wrong is published in full.