How to Tell If an Image or Video Is AI-Generated: 10 Clues and Free Checks

AI-generated photos and deepfake videos are getting harder to spot. Learn the visual clues that still give them away, the free checks anyone can do, and why context matters more than pixels.

Information current as of November 2026 · How we research · Report an error

Magnifying glass on a white desk next to a laptop
Photo by MJ Duford on Unsplash

AI tools can now create photos of events that never happened and videos of people saying things they never said. Some are harmless fun; others are used for scams, fake news and harassment. The uncomfortable truth is that no single trick reliably detects every AI image: the tools improve every few months, and yesterday’s tell-tale signs (like extra fingers) are increasingly rare.

The good news is that you don’t need to be a forensic expert. Combining a few visual checks with a habit of checking the source and context catches most fakes that matter.

Start with context, not pixels

Before zooming in on details, ask the questions professional fact-checkers ask:

  1. Who posted it first? An established news organization, an official account, or an anonymous account created last month?
  2. Is anyone else reporting it? A real dramatic event, such as an explosion, an arrest or a disaster, will be covered by multiple reliable outlets within hours.
  3. Does it trigger a strong emotion? Outrage, fear and “too good to be true” moments are exactly what fake content is designed to provoke.
  4. Is it asking you to do something? Share urgently, donate, invest, or click a link? Be extra careful.

Context checks work no matter how good the AI gets. Visual clues help, but they’re the second line of defense.

Visual clues in AI-generated images

Look closely, ideally on a larger screen, and zoom in.

1. Hands, teeth and ears

Modern generators are much better at hands, but look for fingers that merge, bend unnaturally or vary in number, and for teeth that blur into one strip. Ears and earrings that don’t match are another frequent giveaway.

2. Text and signs

AI images often contain text that looks like writing from a distance but turns into nonsense up close: shop signs, labels, license plates, posters and name badges. Newer models handle short text better, so check small background text especially.

3. Too-perfect skin and lighting

Faces that look airbrushed, glossy or “plastic,” with no pores, blemishes or flyaway hairs, are typical. So is cinematic lighting that’s too perfect for a casual snapshot.

4. Inconsistent light and shadows

Check whether shadows fall in the same direction and whether reflections in windows, glasses and eyes match the scene.

5. Background details that melt

Backgrounds often contain warped railings, impossible architecture, faces in a crowd that dissolve into blobs, or objects that blend into each other.

6. Accessories and patterns

Glasses with mismatched frames, jewelry that fuses into skin, clothing patterns that change halfway across a shirt, and buttons in odd places are common errors.

Clues in deepfake videos

  • Lip sync that’s slightly off, especially on sounds like “p,” “b” and “m.”
  • Blinking that looks unnatural, or eyes that don’t track properly.
  • Blurring or flicker around the edge of the face, hairline or neck, especially when the head turns.
  • Lighting on the face that doesn’t match the room.
  • Voice that sounds flat, lacks breathing, or has an odd rhythm.
  • Low resolution on purpose: fakes are often shared as blurry clips because it hides flaws.

10 free checks anyone can do

  1. Reverse image search. Use Google Lens (search with an image) or TinEye to find where the image appeared first and whether it’s been debunked.
  2. Search for the claim. Type the event plus “fact check” into a search engine.
  3. Check the original account. Look at the profile’s age, previous posts and follower history.
  4. Look for the full-length video. Short clips are easy to manipulate or take out of context.
  5. Check content credentials. Some cameras, editing tools and AI generators attach Content Credentials (based on the C2PA standard), which can show how an image was made. Not all images have them, and their absence proves nothing, but their presence is useful.
  6. Look for AI labels. Major social platforms label some AI-generated content, and some AI companies add invisible watermarks. Labels help, but many fakes slip through unlabeled.
  7. Zoom in on details listed above: hands, text, backgrounds, accessories.
  8. Compare with known real photos of the same person or place.
  9. Check the date and weather. Does the scene match the season, time of day and location claimed?
  10. Wait. For breaking news, waiting a few hours before sharing is one of the most effective habits you can build.

About “AI detector” websites: online detectors can give a useful hint, but they make mistakes in both directions. Treat their result as one signal, not proof.

Why it matters

AI-generated content is used for:

  • Scams, such as fake celebrity endorsements for investment schemes, or fake photos to support “emergency” requests for money.
  • Misinformation during elections, wars and disasters.
  • Harassment, including fake intimate images. In many countries this is illegal; report it to the platform and, where appropriate, to the police.

→ AI Voice Cloning Scams: How They Work and How to Protect Your Family → How to Spot a Phishing Email: 10 Warning Signs

Don’t fall into the opposite trap

As fakes spread, some people dismiss real evidence by claiming it’s AI. Treat “it’s fake” claims with the same care: check the source and look for independent confirmation in both directions.

The bottom line

You can’t rely on spotting extra fingers anymore. Start with context: who posted it, whether reliable sources confirm it, and whether it’s pushing you to act fast. Then use visual clues and a quick reverse image search. When in doubt, don’t share.

→ Want to know how these images are made? Read How AI Image Generators Work.

AI imagesdeepfakesmisinformationfact-checking