All guides
How-to5m readUpdated · 2026-02-10

How to spot a manipulated photo

Photo manipulation predates AI by decades. Here are the classic and modern tells that still work in 2026.

Photo manipulation is older than digital photography itself and remains the most common form of visual misinformation. AI image generation gets the headlines, but old-fashioned Photoshop edits — cropping, splicing, recoloring, removing — are still the bigger volume problem.

Classic manipulation tells

Inconsistent lighting. Every object in a real photo is lit by the same light source(s). Look at shadows: do they all point the same direction? Is the light temperature (warm/cool) consistent across the frame? Mismatched lighting is the #1 splice tell.

Compressed edges around an object. When something is cut from one photo and pasted into another, the pasted object often has slightly different JPEG compression artifacts than the surrounding image. Zoom in on the edges — sharp transitions between “clean” and “noisy” pixels suggest a paste.

Crop edges. Photo crops can be more deceptive than edits. A protest photo cropped tight to look bigger; a one-on-one photo cropped to remove a third person. Reverse image search to find the original — uncropped version is often the more honest framing.

EXIF metadata. Right-click → Properties or use `exiftool`. Real photos have camera model, lens, time, GPS (sometimes). Manipulated photos often have either no EXIF data or telltale software signatures (Photoshop, GIMP).

Reverse image search. Yandex, Google Images, TinEye. If the original is older or differently-framed than the version being shared, the version being shared is the manipulation.

Modern AI-edit tells (different from AI-generated)

Photoshop's generative-fill and similar AI-edit tools insert plausible content into a section of a real photo. The new content typically: - Has slightly different texture / grain than the surrounding photo - May contain garbled detail at high zoom - Often shows tell-tale "smooth" patches where the surrounding noise pattern doesn't continue

The honest summary

For any politically-charged photo: ask yourself whether the framing depends on what’s visible OR what might be just outside the frame. If a politician was photographed alongside a controversial figure, the picture is real but the framing matters — was this a 30-second handshake at a long event, or a long meeting? Reverse-search and find the original story.

Prism Cross-Platform Snapshot lets you see how the same photo is being captioned across every platform — useful for spotting cases where the same image is being weaponized different ways.

Related: How to spot AI-generated images · Content provenance · Misinformation.

Keep reading