How to Tell if an Image Has Been Photoshopped

"Is this image real?" is one of those questions that sounds binary and isn't. Images get cropped, brightened, compressed, and re-saved constantly — all technically edits, all usually innocent. What you actually want to know is whether the image was manipulated: content added, removed, or moved in a way that changes what it's evidence of.

Here's a working sequence, from the checks anyone can do in a browser to the ones that need real tooling.

Start with your eyes — but know what they're good at

Human vision is bad at global properties and decent at local inconsistencies. So don't stare at the image hoping to "feel" the fake; hunt specifics:

• Edges. Zoom to 200–400% around any object that matters. Pasted elements often carry a faint halo, an unnaturally crisp boundary, or blur that doesn't match neighboring edges.
• Light and shadow. Check that shadows fall the same direction for every object, and that reflective surfaces reflect what's actually in the scene. Composites get lighting approximately right; approximately is findable.
• Scale and perspective. Objects added from other photos are routinely a few percent off in size or angle for their position in the scene.
• Repeats. Identical texture patches — the same leaves, the same crowd faces, the same carpet twice — are the signature of clone-stamping. (That one's common enough in documents that it gets its own writeup.)

Check the metadata before anything else technical

Image files carry EXIF metadata: camera model, capture time, GPS, and — usefully — editing software fields. Finding "Adobe Photoshop" in the software tag doesn't convict (photographers edit honestly), and absent metadata doesn't acquit (platforms strip EXIF on upload). What you're reading for is story mismatch: a photo claimed as "taken yesterday on a phone" whose metadata says otherwise, an edit timestamp after the claimed capture, camera fields missing where an original-off-the-phone should have them. Metadata is weak evidence alone but excellent at contradicting stories — the same way PDF metadata works on documents.

Reverse image search the whole image — then the parts

Manipulated images are often assembled from findable sources. Run the image through reverse search, and if a specific region matters — a face, a product, a background — crop and search that region separately. Finding the original with different content in the disputed area is the cleanest possible verdict, and it costs two minutes.

Understand what error level analysis shows

ELA re-compresses the image and maps where compression behaves differently — regions saved a different number of times than their surroundings can light up, which is what pasted or retouched areas often are. It's a genuinely useful lens with genuinely famous failure modes: uniform areas confuse it, screenshots and re-uploads muddy it, and untrained eyes see ghosts in it. We wrote a full explainer on reading ELA properly; the summary is lens, not verdict.

Look at the noise

Every camera sensor leaves a noise fingerprint spread evenly across the frame. Pasted regions bring their own noise from a different camera, a different ISO, a different compression history — and heavy retouching removes noise, leaving patches suspiciously smoother than the rest. Noise-consistency analysis needs software, but it's one of the harder signals to fake, because matching noise across sources is genuinely difficult.

Know when you've hit the tooling threshold

Eyes, metadata, and reverse search catch the careless cases. The careful ones — a competent composite, matched lighting, stripped metadata, fresh compression over everything — need forensic tooling: JPEG quantization-history analysis, copy-move matching, splicing localization, noise maps. If a decision with money or legal weight rides on the image, that's not overkill; it's the appropriate tier, and it's the same escalation logic as any other evidence: quick checks to triage, deep analysis to decide.

Frequently asked questions

Can you always tell if an image was photoshopped?

No — and anyone claiming otherwise is selling something. A skilled edit with matched lighting, rebuilt noise, and clean compression history can be genuinely hard to detect from pixels alone. In practice, most fakes aren't that careful, and manipulators rarely control the surrounding evidence: sources findable by reverse search, metadata that contradicts the story, facts that don't reconcile.

Does screenshotting an image remove evidence of editing?

It removes some — original metadata is gone and compression history is reset — which is exactly why a screenshot where an original should exist is itself a flag. It doesn't remove content-level evidence: cloned regions, lighting inconsistencies, and edge artifacts survive the screenshot.

What's the best free way to check an image?

The trio in this guide's first half: zoom and inspect edges and shadows, read the EXIF metadata against the image's claimed story, and reverse-search the image plus suspicious crops. Ten minutes, no special software, and it resolves a surprising share of cases — in both directions.