A photo of a celebrity doing something outrageous. A “news” image of an event that never happened. A too-perfect property listing. AI-generated images are now good enough to fool most people at a glance — which makes knowing how to spot AI-generated fake images an essential modern skill, not a geeky hobby.

Here’s what to look for, from the obvious glitches to the subtle tells, plus how to verify when it really matters.

How to Spot AI-Generated Fake Images: The 12 Telltale Signs

1. Hands. The classic giveaway. Count the fingers, check how they bend, look at how hands interact with objects. AI still struggles here more than anywhere.
2. Teeth and ears. Zoom in: teeth that merge together, ears at different heights or with odd shapes.
3. Text and logos. Signs, labels, book spines, and logos dissolve into gibberish lettering. Real photos have readable text.
4. Background weirdness. Blurry faces in crowds, people melting into each other, architecture that doesn’t quite connect, railings that go nowhere.
5. Too-perfect skin. Plastic-smooth skin with no pores, especially combined with hyper-detailed everything else.
6. Inconsistent lighting. Shadows falling in different directions, reflections that don’t match the scene, earrings reflecting different rooms.
7. Repeating patterns. AI loves to tile: identical windows, bricks, or leaves repeating in unnatural ways.
8. Overly glossy “AI look”. With experience you develop an eye for it — a certain waxy, hyperreal sheen common to generated images.
9. Impossible physics. Staircases to nowhere, chairs missing legs, glasses with three arms.
10. Metadata absence. Real phone photos carry EXIF data (camera model, date). Generated images typically have none — though this can be faked too.
11. Reverse image search finds nothing. If an image supposedly shows a major news event but appears nowhere else online, be suspicious.
12. The source. Who posted it? A brand-new account with no history sharing “breaking news” photos is a red flag regardless of the image itself.

Verification Tools That Help

• Google reverse image search / TinEye. Check whether the image (or similar versions) already exists online — fakes often reuse elements.
• AI detectors. Tools claiming to detect AI images exist but are unreliable — treat them as one weak signal, not proof.
• Check the metadata. Tools like FotoForensics show compression patterns; inconsistent error levels across an image suggest manipulation.
• Go to primary sources. For news images, check whether reputable outlets (e.g., BBC News) are carrying the story with the same visuals.

Why This Matters Beyond Curiosity

Fake images aren’t just internet pranks. They’re used in:

• Scams — fake property listings, romance scams, fraudulent product photos.
• Disinformation — fabricated “evidence” for political or financial manipulation.
• Fraud — fake IDs, insurance claims with staged damage.
• Non-consensual imagery — a serious and growing harm, particularly targeting women and girls.

Related: deepfake technology and how to spot it covers AI-generated video, which is the next frontier of the same problem.

Teaching Kids to Be Sceptical

Young people encounter AI images constantly on social media. Make “is this real?” a dinner-table game: pull up suspicious images together and hunt for the tells. Media literacy is now as important as road safety — and far less taught. See the risks of AI for children for the broader picture.

The 60-Second Verification Checklist for Viral Images

When an image makes you gasp, angry, or eager to share, run this quick check before believing it:

1. Pause. Emotional reaction is the vulnerability fakes exploit. The pause is the whole defence.
2. Check the source. Who posted it? When was the account created? Do reputable outlets carry the same image?
3. Reverse-search it. Google Lens or TinEye — is this image (or its elements) already online in another context?
4. Zoom in. Hands, text, background faces, edges — 10 seconds of scrutiny catches most fakes.
5. Ask “does this make sense?” Would this person be there, doing that, photographed like this? Context kills most fakes.

If it fails any step, don’t share it. Sharing an unverified shocking image makes you part of the distribution chain — and “I didn’t know” doesn’t undo the harm to whoever’s depicted.

What’s Being Done About It

• Content credentials (C2PA). An industry standard for embedding tamper-evident provenance in images — “this photo was taken by this camera at this time.” Camera makers and platforms are adopting it; it’s the most promising long-term fix.
• Platform labelling. Major social platforms now label some AI-generated content — imperfect, but improving.
• UK law. Creating or sharing non-consensual intimate deepfakes is criminal; the Online Safety Act puts duties on platforms around harmful synthetic media. Enforcement is ramping up.
• Detection research. An ongoing arms race — detectors improve, generators improve faster. Don’t rely on detection alone.

The deeper skill: provenance thinking

Long-term, the skill isn’t spotting fakes — it’s weighing evidence. Who’s showing me this? What’s their track record? Is there corroboration? This is old-fashioned media literacy applied to a new problem, and it’s the durable defence as the technology outruns telltale glitches. Teach it explicitly: to kids, to parents, to colleagues. The fakes will keep improving; critical habits compound.

Fake Images and Scams: Protecting Your Money

AI images power some of the most effective scams targeting UK consumers:

• Fake property listings: AI-generated flat photos for rentals that don’t exist — always view in person (or via live video call) before paying deposits.
• Fake products: too-perfect product shots on scam storefronts. Check the seller’s history, reviews, and whether the business is verifiable.
• Romance scams: AI-generated profile photos of people who don’t exist. Reverse-image-search profile pictures early in any online relationship.
• Investment scams: fake celebrity endorsements with AI-generated imagery. Celebrities don’t DM you about crypto — ever.

The defence is always the same: verify identity, verify existence, never send money under pressure. Report scams to Action Fraud — your report helps protect the next target.

Practical Takeaway

Spotting AI-generated fake images comes down to a habit, not a tool: pause before believing or sharing, zoom in on hands/text/backgrounds, reverse-search anything surprising, and check the source. The fakes will keep getting better — today’s tells will fade — so the durable skill is scepticism itself: extraordinary images need ordinary verification. When in doubt, don’t share.