How to Tell if a Photo or Video Is AI-Generated: Content Credentials and Visual Clues
Visual tells are fading, and AI detectors are unreliable. Here's how to check Content Credentials, what the FBI says to look for, and why the source matters most.
A celebrity endorsing a crypto app. A disaster photo asking for donations. A “proof of life” video from someone you met online. Any of these might be made by AI, and the old tricks for spotting fakes are getting weaker.
The FBI warns that criminals use AI-generated images and videos for fake social media profiles, fake ID documents, celebrity-endorsement scams, fraudulent charity appeals and sextortion.[2] This guide shows the checks that still work.
Step 1: Check the source before the pixels
The most reliable question isn’t “does this look fake?” but “who posted this, and can I confirm it elsewhere?” A real major event will be reported by several established outlets. A real company announcement will be on the company’s own site. If the only source is a new account, a forwarded message or an ad, treat the image as unverified, however real it looks.
Step 2: Look for Content Credentials
Content Credentials are provenance records built on an open standard called C2PA. The Content Authenticity Initiative describes them as a kind of “nutrition label” for digital content: when available, they show where a file came from and how it changed.[1]
If an AI tool supports Content Credentials, the record can show that an image was generated with AI. If a photo was taken with a C2PA-enabled camera and edited in supporting software, that history can be captured too.[1]
To check a file, you can drag it into the free Content Credentials Verify page, which shows any credentials and how the content has changed over time.[3] Some apps and sites also show a small “CR” icon you can tap.
Two limits matter. Content Credentials are still rolling out, so many files have none.[3] And they are not meant to declare what is “real”; they report origin and history so you can decide.[1] A missing record is not proof of anything.
Step 3: Look closely, but don’t rely on it
The FBI lists visual clues that can give away AI images and video: distorted hands or feet, unrealistic teeth or eyes, indistinct or irregular faces, odd accessories such as glasses or jewellery, inaccurate shadows, watermarks, lag, voices that don’t match, and unrealistic movement.[2]
These help, but newer tools make fewer mistakes, so a clean-looking image proves nothing. AI detector websites have the same problem: the Content Authenticity Initiative notes their results can be unreliable and that they are in an arms race with bad actors.[1]
How to stay safe
- Verify the source first. Find the same image or claim from a second, trusted outlet before you believe it.
- Check Content Credentials with a free verify tool, and remember that no record doesn’t mean real.[1] [3]
- Scan for the FBI’s red flags: hands, teeth, eyes, shadows, accessories and unnatural movement.[2]
- Never send money or private details because of an image or video alone, especially to someone you’ve only met online.[2]
- Pause before sharing. If you can’t confirm it, don’t forward it.
When it’s personal
If someone uses AI to make a fake image of you, or threatens to, keep the evidence, don’t pay, and report it to the platform and the police. In the US, the FBI asks victims of AI-enabled fraud to report at ic3.gov.[2]
See real examples in Fake AI Police Officers Are Selling Scam ‘Recovery’ Ads and Scammers Are Using AI Fake Receipts to Cheat Sellers and Stores. For voice fakes, read our family safe word guide.
Protect yourself with Privaku
- Privaku Photos (iPhone) keeps your private photos off the public web, so there’s less to misuse: https://apps.apple.com/app/id6821608477
- Privaku AI Shield (Mac) watches what AI agents do on your computer: https://privaku.com/agent-shield/
Sources
1. Content Authenticity Initiative (accessed Oct 28, 2026): “How it works”. https://www.contentauthenticity.org/how-it-works 2. FBI Internet Crime Complaint Center (Dec 3, 2024): “Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud”. https://www.ic3.gov/PSA/2024/PSA241203 3. Content Credentials Verify (accessed Oct 28, 2026): “Inspect content to dig deeper”. https://verify.contentauthenticity.org/
FAQ
What are Content Credentials?
Content Credentials are provenance information based on the open C2PA standard. When present, they can show which app or device made a file, what edits were made, and whether an AI tool generated it.
If an image has no Content Credentials, is it real?
No. Content Credentials are still rolling out and the metadata can be removed, so a missing record does not prove an image is real or fake.
Are AI image detectors reliable?
Not on their own. The Content Authenticity Initiative says detection tools can be unreliable and are in an arms race with bad actors, so use them as one clue, not a verdict.
See the post on YouTube Watch and follow AI Safe Mode on YouTube @AISafeMode and Instagram @aisafemode.
References
- Content Authenticity Initiative, Accessed Oct 28, 2026. How it works. Read the report ↗
- FBI Internet Crime Complaint Center, Dec 3, 2024. Criminals Use Generative Artificial Intelligence to Facilitate Financial Fraud (I-120324-PSA). Read the report ↗
- Content Credentials Verify, Accessed Oct 28, 2026. Inspect content to dig deeper. Read the report ↗
