How to Verify a Viral Video in 30 Seconds

A shocking clip lands in your feed. A politician says something career-ending, a celebrity does something bizarre, a disaster unfolds somewhere far away. You have two options: share it in three seconds, or spend thirty verifying it.

Here is the thirty-second routine. It will not catch everything — nothing does — but it catches most fakes, most of the time.

Step 1: Check the source (10 seconds)

Before you look at a single pixel, look at who posted it.

• How old is the account? Extraordinary footage from an account created last week is the oldest red flag on the internet.
• What else do they post? A feed full of engagement bait, crypto links or one-sided outrage is not a news source.
• Where is the original? If the clip shows a public event, real news agencies would have it within hours. If nobody reputable is carrying it, ask yourself why you are the lucky one seeing it.

One honest question filters half of all fakes: does this video exist anywhere credible, or only in reposts?

Step 2: Reverse-search a frame (10 seconds)

Take a screenshot of a distinctive frame and drop it into Google Lens or a reverse image search.

You are looking for two things:

1. The same clip, earlier, with a different story. A huge share of "breaking" videos are old footage recycled with a new caption — a 2023 explosion relabelled as this week's news. This is technically not even AI fakery, and it is by far the most common kind of visual misinformation.
2. Fact-checks that already exist. If the clip is viral, chances are someone has already debunked it. Search a few keywords from the caption too.

Step 3: Run it through a detector (10 seconds)

Your eyes can be fooled — the best AI clips now survive every visual check, as we showed in our guide to spotting AI-generated video. A detector reads statistical traces in the pixels that generators leave behind and human vision cannot register.

Paste the video link (TikTok, Instagram, YouTube and X links work directly) or upload the file at AI Video Check — no account needed for a quick check. Twenty seconds later you have a probability score and, when possible, a guess at which generator made the clip.

How to read the result:

• AI-generated (90%+) or Probably AI (57–90%): treat the clip as synthetic unless strong evidence says otherwise — the higher the number, the more confident the call.
• No AI signals or Probably not AI (under 57%): the detector found no traces of a generator — which is not the same as "this really happened". Remember steps 1 and 2, because real footage with a false caption is still a lie.
• Scores near the 50/50 middle are the least certain — usually heavy re-compression or something genuinely novel. Fall back on the source.

One limit worth knowing before you lean on any detector, ours included: it looks for the fingerprints of generative models — Sora, Veo, Kling and the rest. It does not judge editing, VFX, CGI or a face swapped onto real footage. A staged clip shot on a phone, or a scene built by a VFX house, can come back clean and still be a lie. "No AI signals" is a statement about how the pixels were made, not about what they show.

The mindset that ties it together

Misinformation does not spread because people are stupid. It spreads because a video confirms what we already feel, and sharing takes less effort than checking.

So make the routine mechanical, not moral: source, reverse search, detector. Ten seconds each. If a clip fails any step, do not share it — or share it with your doubt attached.

Real videos are innocent until proven guilty. Everything extraordinary deserves thirty seconds of skepticism first.