A shocking photo, a celebrity selling something, a video that seems too perfect. Is it real? There's no single test that can tell you for certain, and AI tools keep getting better. But a few checks can tell you a lot, especially about where something came from. Here's what helps, and what its limits are.

⚡ The short version

Find where it came from. Search for the image to see who posted it first and when.
Look for a label, such as YouTube's “Made with AI” or Meta's “AI info.” No label doesn't prove it's real.
Use a checker, like Google's Gemini app or the Content Credentials site. They can confirm some AI images but can't clear others.
If it asks for money or rushes you, treat it as a scam however real it looks.

1. Check where it came from

Google's About this image tool can show when Google first saw the image (or similar ones), other pages that use it, and its earliest uses. An old photo passed off as new, or a picture that first appeared on a joke account, is a red flag.

In the Google app or on the Google homepage, search for the image and open it in the image viewer.
Select About, then Learn more about an image.

Google says this is only available in some regions. When there's information on how an image was made, Google shows it under “How the image was made or edited,” but it warns that this kind of information can be changed with editing tools.

2. Look for an AI label

YouTube: open the full video description and look for “How this content was made.” “Made with AI” means it was meaningfully altered or generated with AI. It's there because the creator disclosed it, YouTube's own AI tools were used, or the file carried Content Credentials. YouTube may also show a label on the video player.
Facebook and Instagram: Meta adds an “AI info” label when it detects industry-standard AI signals or when the person posting says it's AI. For posts it detects were only edited with AI, Meta moved the label into the post's menu.

A missing label proves nothing. These labels depend on people being honest or on signals that AI tools add, and those can be missing or removed.

3. Use a checker, and know its limits

Gemini app (Google): sign in at gemini.google.com, select Add files, upload the image, video or audio, and ask “Is this AI-generated?” Gemini looks for Google's invisible SynthID watermark and for Content Credentials.
- If SynthID is found, Google says all or part of it was made or edited with Google's AI.
- If it isn't found, Google's AI probably didn't make it, though the result can be unclear for small edits or heavily changed files. Another AI tool could have made it.
- One file at a time. Videos must be under 90 seconds. Google mentions a limit of about 10 checks of each type per day.
Content Credentials Verify: at contentcredentials.org/verify, drop in a file to see any history it carries, including whether AI tools were used. Many files have no Content Credentials at all, so “no information” is common and doesn't mean anything either way.

For best results, use the original file, not a screenshot. If you must use a screenshot, Google suggests cropping tightly around the image.

4. Look closely

The FBI and Google list clues that can give AI away. Realistic fakes may show none of them, so use these alongside the checks above, not instead of them:

Hands, feet, teeth or eyes that look wrong, or faces that are blurry or uneven
Glasses or jewelry that don't look right
Shadows and lighting that don't match
Garbled or misspelled text on signs, labels or clothing
Repeating patterns that look unnatural
In video: lag, a voice that doesn't quite match, or movements that look unreal

5. When it's about money, it's a scam until proven otherwise

The FBI says criminals use AI-made images and video for celebrity endorsements of fake products, fake charity appeals after disasters, fake IDs, and “proof” that an online contact is a real person. If a photo or video is pushing you to pay, invest, donate or share personal details, stop and check it with a source you trust. See Is this text, email or link a scam?

More on AI: Turn off or limit AI in your apps · Voice-clone scams and a family safe word · Is that photo or video AI?

Is that photo or video AI? Checks that actually help