How to Tell if an Image Is AI-Generated (and Spot Deepfake Video and Audio)

Short answer: You usually can't tell by looking alone, so check where the image came from instead. Run a reverse image search, look for Content Credentials, and ask Google's Gemini whether it carries Google's SynthID watermark. If the source is unclear and nothing confirms it, treat the image as unverified.

This guide is for anyone who wants to check a viral photo, a suspicious video call or a fake of themselves.

Digital forensics expert Hany Farid shares practical tips for judging whether a striking online image is real in this TED talk.

What's the quickest way to check a photo?

Work through these five checks in order. Most fakes fall apart at step 1 or 2.

1. Find the source: who first posted it, and is that account or outlet trustworthy?
2. Reverse-search it: in Chrome, right-click the image and choose Search with Google Lens, or try TinEye.
3. Check Content Credentials: upload the file at contentcredentials.org/verify.
4. Check for Google's watermark: upload it to Gemini and ask "Is this AI-generated?"
5. Look for context: do the weather, signs, clothing and news reports match the claimed time and place?

Reverse search shows where else the picture appears and whether an older version exists. Google's help page says About this image also shows when Google may first have seen an image and, where available, whether it was captured with a camera or generated with AI. Google notes the feature is only available in some regions.

How do you check Content Credentials and AI labels?

Content Credentials (the C2PA standard) are signed records of how a file was created and edited. Google describes them as showing how media was made "with or without AI."

To check for Google's SynthID watermark in Gemini on a computer:

1. Go to gemini.google.com and sign in.
2. Click Add files, then Upload Files.
3. Choose one image (100 MB or less).
4. Ask "Is this AI-generated?"

For screenshots, Google advises cropping tightly around the image and not uploading a collage. As of November 2026, the quota is about 10 image checks in any 24 hours. Gemini also reads Content Credentials on the web and Android app.

Did you know? A "no watermark" result from Gemini doesn't mean a photo is real. Google says it only means the image "wasn't created or edited by Google AI." It "could have been created by other AI systems."

Metadata can lie too. Google's Search help warns that AI-related metadata "can be modified with image generation and editing tools, so it may not be accurate." A missing label proves nothing; a present one is a clue, not a verdict.

Can you still spot AI images by looking?

Sometimes. The FBI lists tells such as distorted hands or feet, unrealistic teeth or eyes, irregular faces, odd glasses or jewellery, and inaccurate shadows.

Be honest with yourself about the limits. Newer image models fix many of these flaws, and real photos can have odd shadows or blur too. In our view, visual tells should raise suspicion, never settle the question.

Do free AI image detectors work?

They can help, but no detector is reliable on its own. Every detector gives a probability, and in our view you should expect both false positives (real photos flagged as AI) and misses.

Hive's free Chrome extension, from Castle Global Inc., scores images, video, text and audio for AI generation. It also guesses which generator made an image.

Uploading a photo sends it to that company. TinEye says it never saves uploaded images. For Gemini and Hive, read their privacy pages before uploading anything private.

Never accuse a person based on a detector score alone. If you're curious how these images are made in the first place, see our guide to creating AI images for free, and our explainer on animating photos with AI for moving fakes.

How do you spot deepfake audio and video calls?

On a live call, verification beats detection. The FBI advises families to agree a secret word and to hang up and call back on a known number.

1. Don't act on any request for money or passwords during the call.
2. Ask for your family safe word, or a question only the real person could answer.
3. Hang up.
4. Call back on a number you already have, not one they give you.
5. Never send money, gift cards or cryptocurrency to someone you know only from a call.

For voice notes and calls, the FBI suggests listening closely to tone and word choice. Our guide to AI voice cloning explains how little audio a cloner needs, and how to protect your family.

What should you do if someone made a deepfake of you?

Act quickly, but save evidence first.

1. Screenshot the post and copy its link, username and date.
2. Report it on the platform. On YouTube, use the privacy complaint for AI content that "looks or sounds like you."
3. For intimate images, use StopNCII.org if you're over 18 and were 18 or older when the image was taken.
4. In India, report it at cybercrime.gov.in; for financial fraud, also call 1930.
5. Tell people you trust, so they hear about it from you first.

StopNCII creates a hash (a digital fingerprint) of the image on your device. Only the hash is shared with participating companies, never the photo itself.

For broader protection tools, browse our security tools category.

Frequently Asked Questions

How do I check if a photo is AI-generated?

Find the original source, reverse-search the image, and check Content Credentials and Google's SynthID watermark. If nothing confirms where it came from, treat it as unverified.

Can metadata tell me if an image is AI-generated?

Sometimes. Content Credentials or AI-source metadata can show AI was used, but Google warns such metadata can be edited, and many images have none.

Is there a free AI image detector I can use online?

As of November 2026, Hive's Chrome extension is free with no login, and Gemini's SynthID check has a daily quota. Neither is proof on its own, so combine them with source checks.

How do I detect a deepfake video call?

Don't rely on spotting glitches. Ask for a safe word, hang up, and call the person back on a number you already know.

How can I tell if an image is real?

A trustworthy original source, an older matching version from reverse search, and supporting news coverage are your best evidence.

Sources

• TED: Hany Farid, How to spot fake AI photos
• Google Search Help: Search with an image
• Google Search Help: Learn more about an image
• Gemini Apps Help: Verify AI-generated images, videos and audio
• Google: Making it easier to understand how content was created and edited
• Content Credentials Verify
• TinEye help: Does TinEye keep images I upload?
• Hive AI Detector on the Chrome Web Store
• FBI IC3: Criminals use generative AI to facilitate financial fraud
• YouTube Help: Protecting your identity
• StopNCII.org
• I4C (Ministry of Home Affairs): National Cyber Crime Reporting Portal and 1930 helpline

Reviewed on November 3, 2026. Google is expanding SynthID and Content Credentials checks to Search and Chrome, so the easiest place to check may change.