How will we know what's real when AI can create convincing fake videos?

We probably won’t be able to judge authenticity by looking or listening alone. The answer will be a combination of **provenance, independent verification, and healthy skepticism**—not a single “deepfake detector.”

For important footage, check:

- **Where it came from:** Is there an original recording, and can its source and history be traced? Tools such as cryptographic signatures or content credentials can help show how a file was created or edited, though they don’t prove that the recording itself tells the whole truth.
- **Who else can confirm it:** Look for independent reporting, witnesses, documents, or other recordings. Several outlets repeating the same unverified clip are not necessarily independent confirmation.
- **What the context is:** Check the date, location, caption, and whether the video may be edited or presented out of context. A genuine recording can still be misleading.
- **How reliable the claim is:** Be cautious with sensational clips, especially when the source is unknown or the claim is still developing. Waiting for verification is often wiser than sharing immediately.

AI-detection tools may offer clues, but they can make mistakes and will have to keep up with new techniques. Watermarks and metadata can also be removed or forged. So for consequential claims—such as evidence of a crime, a public emergency, or a statement attributed to someone—seek confirmation from trusted, accountable sources rather than relying on a clip alone.

The goal won’t be to trust every video. It will be to ask how it was verified, by whom, and what evidence supports it.