How can you tell if a video was made with AI?
Short answer
There is no definitive way. Manual checks and tools give you clues, not proof. As generation tools improve, visible mistakes become rarer; and real videos can show flaws that look "synthetic" because of compression, filters or low light. The most reliable approach is to look less at the pixels and more at where the video came from.
1. Go to the source
- Who posted the video first? Is the account you saw the original source, or a re-upload?
- Does the caption or the comments say it was made with AI? Some platforms show a label on such content. No label does not mean the video is real.
- What does the account usually post? A steady stream of "unbelievable" clips is a warning sign.
2. Look for the event elsewhere
- If the video shows something significant, news outlets and official accounts will mention it too. If nobody else does, be careful.
- Take a frame from the video and run a reverse image search. An older version, or the same footage in a different context, may turn up.
3. Look closely at the picture
These are clues; none of them is conclusive on its own.
- Hands and fingers: their number, shape, and how they hold objects.
- Text: are signs, plates or labels garbled or meaningless?
- Edges: melting or flicker around hair, glasses or ears?
- Physics: do shadows, reflections, water and smoke behave naturally?
- Continuity: when the camera moves or the scene cuts, do objects, clothes and background stay the same?
- Background: are faces and movement in a crowd consistent?
4. Listen to the audio
- Do the lip movements match the speech?
- Does the sound fit the setting (room echo, outdoor noise, breathing)?
- If it is a familiar person's voice, does what they say appear in their other recordings?
5. Separate the footage from the claim
An AI-generated video can carry accurate information, and a real video can carry a false claim. "Is this footage real?" and "Is what it says true?" are two different questions.
What does Fact AI do here?
Fact AI looks for signs of AI generation in the video or image and shows a likelihood percentage, with the signs that led to it. It does not give a definitive "AI or not" verdict. In compilation videos it separates which parts look synthetic and which look real. When it cannot assess the footage it says so plainly; that does not mean "real".
This reading is separate from checking the video's claims against sources: the same analysis answers both questions separately. Fact AI can be wrong too; use the steps above together with its reading. How an analysis is made: Method.