When a video goes viral during breaking news, the reflex is to ask whether it's AI-generated. Usually it isn't. The Global Investigative Journalism Network and Bellingcat both make the same point in their verification guides: the dominant form of video misinformation is far more mundane and far more effective — genuine footage lifted from one event and re-shared with a caption tying it to another. A real fire, a real protest, a real flood — just from the wrong country or the wrong year.
That's good news, because you don't need forensic tools to catch it. You need to find where the clip first appeared and what it originally showed. If you're instead worried the footage itself is synthetic, that's a different check — see our guide on telling if a video is AI-generated.
The verification workflow
Grab a few keyframes
You can't reverse-search a whole video, so pull still frames from it first. The InVID/WeVerify browser plugin does this automatically — right-click and choose 'Analyse video', and it fragments the clip into distinctive keyframes. No plugin handy? Pause on a frame with something specific in it (a storefront, a license plate, a skyline) and screenshot it. Those are the frames worth searching.
Reverse-search each frame — and sort by oldest
Drop each keyframe into Google Lens, TinEye, and Yandex. Then do the one thing most people skip: sort by date. TinEye lets you order results 'oldest first', which surfaces the earliest time that frame was indexed. A clip captioned 'happening right now' that first showed up online three years ago is debunked on the timestamp alone — you don't even need to watch it.
Read the frame, not the caption
Authentic footage carries its own date and place clues. Look at the weather, the state of the trees (bare, green, autumn), the length of shadows, visible signage and its language, license-plate formats, and architecture. If a caption says 'Gaza today' but the shop signs are in Portuguese and the trees are in full summer leaf during a claimed winter event, the caption is doing the lying, not the video.
Find the first upload, then read its original caption
Context stripping — real footage with a fabricated caption — is the single most common video hoax, far more common than deepfakes. Once reverse search points you to the earliest post, read how it was originally described. Recycled protest, disaster, and war footage almost always traces back to a different country, a different year, or a completely different event.
Cross-check before you trust it
Search the claim on Google Fact Check Explorer and run it through FAXTR, which aggregates 100+ IFCN-signatory fact-checkers in one query. Viral recycled clips are often debunked within hours, so the verdict may already exist. If two independent checkers trace the video to the same original context, treat that as settled.
Clues hiding inside the frame
Before you reach for a single tool, look closely at what the video actually shows. Fact-checkers call this reading the frame — and it often settles the question on its own.
Season and vegetation
Bare branches, green canopy, or autumn color pin the footage to a time of year — and expose captions that claim the wrong season.
Signage and language
Shop names, road signs, and their alphabet reveal the country. Footage 'from' one place with signs in another language rarely survives this check.
Weather and light
Wet streets during a claimed drought, or long shadows during a 'midday' event, are quiet contradictions worth chasing.
Plates, uniforms, currency
License-plate formats, police or military uniforms, and visible banknotes narrow down the location fast and are hard to fake convincingly.
Tools worth having
InVID / WeVerify
The standard browser plugin for journalists and human-rights researchers. Auto-extracts keyframes, runs them through multiple engines, and reads available metadata.
TinEye
Reverse image search with a crucial 'oldest first' sort — the fastest way to catch a clip that's older than its caption claims.
Google Lens
Broadest match database. Best for identifying locations, buildings, and objects inside a frame.
Yandex Images
Often outperforms others on faces and Eastern European or Middle Eastern scenes — worth running when Google and TinEye come up empty.
Google Fact Check Explorer
Searches published fact-checks directly. If a viral clip has already been debunked, its verdict frequently shows up here first.
A caution on false matches
Reverse search is a lead, not a verdict. Bellingcat's own toolkit warns that partial metadata and coincidental visual matches can point you the wrong way — two different fires can look almost identical in a low-resolution frame. Treat any single match as a hypothesis and confirm it with a second, independent signal before you decide: a matching landmark, a corroborating report, a fact-check that reached the same origin.
If it turns out to be recycled
Don't re-upload the clip to debunk it — that just feeds it more reach. Reply with a link to the original context or the fact-check instead, and report the post through the platform's misinformation flow. Describe the footage, point to where it really came from, and let the miscaptioned version lose momentum on its own.
Check a viral claim in seconds
FAXTR searches 100+ fact-checking organizations in one query — free, no login required.