How to Tell if an Image Is AI-Generated

To tell if an image is AI-generated, look for broken details that generators still struggle with — hands and fingers, text and logos, eyes and teeth, warped backgrounds, and lighting that doesn't add up. No single sign is proof, so weigh several together and, when it matters, verify the source.

Key takeaways

The tell-tale signs of an AI image

1. Hands and fingers

Hands are still the classic giveaway. Look for the wrong number of fingers, fingers that bend oddly or merge, and hands that grip objects that aren't quite there. The same applies to feet and to hands interacting with other hands.

2. Text and logos

AI models often render text as gibberish — letters that almost spell something, warped signage, or brand logos that are subtly wrong. Zoom into any writing in the image: real photos have consistent, legible text; AI images frequently don't.

3. Eyes, teeth and ears

Faces can look convincing at a glance but fall apart in the details: mismatched eyes, irregular pupils, too many or oddly shaped teeth, and asymmetrical or melting ears. Reflections in the eyes that don't match the scene are another clue.

4. Backgrounds and repeating patterns

Generators focus detail on the subject and get lazy in the background. Watch for warped architecture, objects that blur into each other, fences or railings that don't line up, and textures (leaves, crowds, tiles) that repeat unnaturally.

5. Lighting, shadows and reflections

Physics is hard to fake. Shadows falling in different directions, reflections that don't match the object, and light sources that don't explain the highlights are all signs the image was generated rather than captured.

6. Skin, hair and materials

AI skin is often too smooth or waxy; hair can blend into the background or have strands that go nowhere; jewelry, glasses and fabric may have details that don't quite connect. Overall, an image that looks too clean and symmetrical is worth a second look.

Why spotting AI images is getting harder

Each generation of image models fixes the last generation's mistakes. Hands are far better than they were, text is improving, and photorealistic lighting is increasingly convincing. That means a checklist alone will fail more often over time — and it's exactly why building an intuition through practice matters more than memorizing signs.

What about AI detectors?

Automated detectors analyze an image and return a likelihood that it's AI-generated. They're a useful second opinion, but they're not proof: they produce false positives (flagging real photos) and false negatives (missing fakes), and their accuracy drops on screenshots, crops and re-compressed images. Use them alongside your own judgment and the source of the image, not instead of them.

Train your eye — play Real or AI?

A free endless game: two images, one real and one AI. Tap the fake and learn instantly. It's the fastest way to build the intuition this guide describes.

Play Real or AI? →

Frequently asked questions

Can you always tell if an image is AI-generated?

No. The best 2026 image models can produce photos with no obvious flaws. Visual tells like broken hands or garbled text still catch many fakes, but they're disappearing fast, so no single sign is proof on its own.

What are the most common signs of an AI image?

The most reliable tells are malformed hands and fingers, garbled text and logos, asymmetrical or unnatural eyes and teeth, warped backgrounds, inconsistent lighting and shadows, and skin or textures that look too smooth or too repetitive.

Do AI image detectors work?

AI detection tools give a probability, not certainty. They can help as a second opinion but produce both false positives and false negatives — especially on compressed, cropped or edited images — so they shouldn't be treated as definitive proof.

How can I get better at spotting AI images?

Practice with feedback. Playing a real-versus-AI game where you guess and immediately learn the answer trains your eye far faster than a checklist, because you build an intuition for the subtle patterns models leave behind.