How to Spot an AI-Generated Photo: 10 Clues

A photo pops up in your feed: a jaw-dropping storm over a famous city, a “rare” animal nobody has heard of, a dramatic moment from a breaking news story. It looks real. It has thousands of shares. And something about it feels slightly too perfect.

Learning how to tell if an image is AI generated is now a basic internet skill, like spotting a phishing email. AI images have gotten dramatically better, and the old “count the fingers” trick doesn’t always work anymore. But there are still plenty of tells if you know where to look.

Here are 10 clues, from quick visual checks to a little detective work, plus an honest look at what no checklist can promise.

How to tell if an image is AI generated: start by zooming in

AI image generators are excellent at the overall impression and weaker at small, logical details. A picture can look flawless on a phone screen and fall apart at 300% zoom. So the first rule is simple: look closer before you react.

1. Hands and fingers

Hands are complex, constantly changing shape and often partly hidden, which made them notoriously hard for AI. Newer models are much better, but check anyway: count fingers, look for fused or bending-the-wrong-way joints, and notice hands that merge into objects like cups or phones.

2. Text, signs and logos

Look at street signs, shop names, T-shirt prints, book spines and license plates. AI often produces letters that look like a language from a distance but turn into gibberish up close, or words that are almost right. Some newer tools render text well, so treat clean text as neutral, not as proof of a real photo.

3. Eyes, teeth, ears and jewelry

Real faces have small asymmetries, but AI faces can have odd ones: pupils of different shapes, a row of too many teeth, ears that don’t match, earrings that differ from one side to the other or melt into the earlobe. Glasses are another giveaway when the frames don’t connect properly.

4. Skin that’s too perfect

Many AI portraits have a polished, almost waxy smoothness, with no pores, flyaway hairs or uneven texture. Everyone looks like they’re lit for a magazine cover. Real people, even in good photos, usually have a little visual “noise.”

5. Hair and edges

Check where hair meets the background, where a shoulder meets a wall, or where a hat meets a head. AI sometimes blurs these boundaries into a soft smear, or lets strands of hair dissolve into the scenery.

Then look at the whole scene

Once you’ve checked the details, zoom back out and ask whether the scene makes physical sense.

6. Backgrounds that don’t hold up

The main subject usually gets the AI’s best effort; the background gets less. Look for warped railings, windows that don’t line up, staircases going nowhere, cars melting into each other, and crowds where faces turn into blobs.

7. Light and shadows

Find the main light source, then check whether shadows fall consistently away from it. AI images may light a face from the left while shadows on the ground point in another direction, or show a subject brightly lit in a scene that’s otherwise dim.

8. Reflections and physics

Mirrors, windows, water and sunglasses are tricky. A reflection might show a different pose, or nothing at all. Also notice objects that defy physics: a cup floating slightly above a table, a strap that goes through a shoulder, fabric that blends into skin.

Finally, check the context

Visual clues are getting harder to find as models improve. Context clues are often more reliable, and they work for edited real photos too.

9. The source and the story

• Who posted it first? Scroll back to the original account. Is it a known photographer or outlet, or a brand-new page posting viral content?
• Do trusted sources show the same thing? A dramatic news event will usually be covered by multiple outlets from several angles.
• Try a reverse image search. Tools like Google Lens or TinEye can show where an image has appeared before and whether an earlier, different version exists.
• Read the comments carefully. Sometimes someone has already found the original or pointed out the flaw.
• Notice your emotions. Images designed to make you furious or amazed spread fastest. That’s exactly when to slow down.

10. Labels, watermarks and Content Credentials

Some AI tools add visible watermarks, invisible watermarks or metadata that records how an image was made. Some social platforms show “AI-generated” or “made with AI” labels based on that information.

One open standard for this is Content Credentials. You can upload an image to the free Content Credentials Verify tool to see whether it carries a record of its origin and edits. Just remember: metadata is often stripped when images are screenshotted or reposted, so “no credentials” doesn’t mean “real.”

Quick checklist: real photo or AI?

What about AI detector websites?

There are many online tools that claim to tell you whether an image is AI. They can be a useful extra signal, but they’re not a verdict. Results can disagree from one tool to the next, and both false alarms and misses happen, especially with compressed, cropped or edited images.

Use detectors the way you’d use a second opinion: helpful when it agrees with other evidence, not enough on its own to accuse someone of faking a photo.

Edited, enhanced or generated? It’s a spectrum

Not every AI-touched image is a fake. Phones now use AI to brighten night shots. People remove photobombers with free AI photo editors, sharpen family photos when they restore old pictures with AI, or play with AI face filters for fun.

A useful question isn’t just “Is this AI?” but “Is this image presented honestly?” A clearly labeled aging filter is fine. A fully generated “news photo” passed off as real is not.

Train your eye in five minutes a week

Spotting fakes gets easier with practice. When a news outlet or fact-checker debunks a viral AI image, open it and look for the clues yourself before reading their explanation. You’ll start noticing patterns fast.

Another fun exercise: make a few AI images yourself with a free tool, then study them at full zoom. Seeing where the generator struggles with your own prompts is one of the quickest ways to build intuition. It’s also a good conversation starter with kids and older relatives, who are often the most targeted by fake images.

What to do when you find a fake

1. Don’t share it, even to mock it. Shares boost reach either way.
2. Report it using the platform’s tools for false or manipulated media.
3. Correct gently. If a friend posted it, a private message with your reasons lands better than a public pile-on.
4. Protect yourself. If someone creates fake images of you, save evidence and report it. Our guide on whether it’s safe to upload your face to AI apps covers steps to take.

FAQ

Can AI images be detected 100% of the time?

No. The best AI images can fool both people and detection tools, and the technology keeps improving. Combining visual clues, context checks and metadata gives you the best odds, but some uncertainty will always remain.

Do AI images still get hands wrong?

Less often than before. Hands remain worth checking, but newer models frequently draw them correctly. Look at several areas, including text, jewelry, backgrounds and shadows, rather than relying on one clue.

How can I check if a photo is AI on my phone?

Zoom in on details, run a reverse image search with a tool like Google Lens, look for platform labels such as “AI info,” and upload the image to a Content Credentials checker. Then see whether trusted sources report the same scene.

Is it illegal to post AI-generated images?

Generally, posting AI art or fun edits is legal. Problems arise when images are used to deceive, defame, harass or impersonate someone, and rules vary by place and platform. When in doubt, label AI images clearly.

Spotting AI photos is less about having a perfect eye and more about building a habit: zoom in, zoom out, check the source, and pause before you share. Do that, and you’ll catch most fakes, and avoid passing them along when you can’t be sure.