Can AI edits always be detected?

Not from the picture alone. Good AI edits can leave no visible trace, and detectors make mistakes. Certainty needs proof recorded when the photo was taken, like a Timesnap fingerprint.

Does an AI eraser leave a trace?

Sometimes: a smeared or repeated patch of background where the object was. On a plain or busy background it can be invisible. A Timesnap photo checked against its fingerprint shows the erased area in red.

Do phones label AI-edited photos?

Some editing tools add a label or Content Credentials to photos they change, but not all do, and labels are often lost when a photo is shared. Don't rely on a missing label.

How to Spot an AI-Generated or AI-Edited Photo

Phone apps can now erase people, add objects and change whole scenes in seconds. Here's what gives AI away, and what to do when nothing does.

AI image tools come in two kinds. Generators make a whole picture from a text prompt. Editors change part of a real photo: eraser tools remove things, generative fill adds or replaces them, and some apps change faces, signs or the sky. The second kind is harder to catch, because most of the photo is genuinely real.

Signs of an AI-made or AI-changed photo

Hands, teeth and ears: extra or merged fingers, too many teeth, earrings that don't match. Newer tools are better at these, but mistakes still happen.
Text: signs, labels and number plates with letters that look almost right but don't spell anything, or change style halfway.
Patterns that melt: fences, tiles, brickwork and railings that bend, merge or stop making sense in the background.
Too smooth: skin, walls or surfaces with no grain at all, especially next to areas that do have grain.
Light that doesn't add up: shadows missing or falling the wrong way, reflections that don't match the scene.
Patched areas: where something was erased, the background can look smeared or slightly repeated, often in a shape the size of the removed object.

None of these signs proves anything on its own, and the best AI images show none of them. Use them as reasons to look further.

Look for Content Credentials

Some AI tools, cameras and editing apps now attach Content Credentials (the C2PA standard) to the images they make or change, recording which tools were used. You can check a photo for them at contentcredentials.org/verify. If the credentials say an AI tool was used, that's strong evidence. If there are none, it tells you nothing: most photos don't carry them yet, and they're lost when a photo is screenshotted or sent through most apps.

Search for the source

Run the image through a reverse image search such as Google Lens or TinEye. AI pictures that went viral have often already been checked by fact-checkers, and a search can turn up the real photo an edit was made from.

Be careful with AI detectors

Online "AI or not" detectors give a percentage score, but they are often wrong: real photos can score as fake, especially edited or compressed ones, and new AI tools can fool detectors trained on older ones. A detector's score is a hint, never a verdict, and it can't tell you which part of a photo was changed.

The reliable answer: proof from the moment of capture

Spotting AI after the fact gets harder every year. The dependable way round it is to record proof when the photo is taken. The Timesnap camera stamps the real time and place and stores a fingerprint of the photo, about 50,000 tiny squares, on our servers, where nobody can change it, not even the photographer. Later, anyone can drop a copy on timesnap.info and see it compared square by square. If AI has erased, added or changed something, that area is outlined in red.

For work where photos are evidence, such as insurance, property, construction, security and deliveries, asking for photos taken this way removes the guesswork.

Take photos with the real time and place on them: get the free Timesnap camera.

How to spot an AI-generated or AI-edited photo