Verify a Fake Video: A Practical Way to Check What You’re Seeing
About this check
Start with the claim, not the clip
To verify a fake video, first write down the specific claim the video is supposed to prove. Is it claiming that a person said something, that an event happened, or that a recording was made in a particular place and time? A finished check should answer that claim, show the evidence behind the answer, and make clear when the available evidence is not enough.
Do not treat “real video” and “true claim” as the same question. A recording can be genuine while its caption is wrong, outdated, or missing context. Check the statement attached to the video as well as the media itself.
A manual method for checking a video
1. Preserve what you saw
Save the link, the account that posted it, the caption, and the date you encountered it. If the post may change or disappear, keep a screenshot that includes the surrounding text. Note whether the video is presented as original footage, a repost, a reaction, or a recording of another screen.
This record gives you something specific to compare. It also prevents a later search from quietly changing the question from “Did this event happen?” to the narrower question “Does this clip exist online?”
2. Separate the visible details from the interpretation
Describe only what can be observed: people, signs, spoken words, logos, weather, landmarks, subtitles, and other text. Then write the interpretation separately. For example, “a person is speaking in front of a building” is different from “this is a speech at a particular event.”
Look for frames that contain distinctive text, faces, signs, or locations. Pausing the video and examining several frames can give you search terms. If the video includes speech, write down an exact phrase and search it in quotation marks, while also trying shorter combinations of the most distinctive words.
3. Search for earlier context
Search the exact caption, unusual phrases, names, locations, and dates. Search both the claim and neutral descriptions of what appears in the video. A reverse-image search of a clear frame may also reveal earlier uses of the same image or related material.
Treat search results as leads, not proof. A result may repeat the original post, show only a cropped version, or omit the context you need. Reposts can also preserve a clip while changing its caption. Compare dates, locations, wording, and the earliest version you can find rather than counting the number of accounts repeating it.
4. Compare independent evidence
Look for reporting, official records, full-length recordings, direct statements, or other material that addresses the same event. Check who produced each source, what evidence it provides, and whether it is reporting an observation or repeating someone else’s conclusion.
A source can be relevant without settling the claim. Give more weight to material that shows its underlying evidence and less to a post that offers only certainty, outrage, or a label. If the sources conflict, record the conflict instead of selecting the most confident caption.
5. Inspect the media, without overclaiming
Watch for abrupt cuts, mismatched audio, unusual movement, inconsistent lighting, distorted text, or faces that change between frames. These details can justify closer examination, but they do not by themselves prove that a video is AI-generated or deliberately manipulated. Compression, editing, poor lighting, and screen recordings can also make footage difficult to interpret.
That is where a manual check may stop working. You may not find an original upload. Search engines may surface only copies. A short clip may leave too little context, and available sources may be unreliable or disagree. In those cases, the responsible result is not a forced yes or no. State what is supported, what remains uncertain, and which missing evidence would change the assessment.
What VOM adds to the check
VOM accepts voice input for submitting a claim to verify. It answers a checked claim with a verdict, a confidence score, and any sources it found. Its verdict can be true, false, or inconclusive, so a claim the evidence cannot settle is reported as inconclusive rather than forced into a yes or no.
VOM checks claims using real-time web search, can reuse a closely matching earlier check when a claim is not time-sensitive, and shows the sources it found. It can also warn you when an uploaded image appears AI-generated and read the text inside a screenshot, so a forwarded image can be checked without retyping it.
VOM independently reviews an AI-generated image or video before approving it for publication. Separately, its article-sharing pipeline still publishes if its AI image or video generation is blocked, fails, or is exhausted.