How to Tell if an Image Is AI Generated

To tell if an image is AI generated, zoom in on the hands, read every visible word, and check whether the light, shadows, and reflections agree. An odd detail is a reason to investigate, not proof by itself. If you have the file, you can also check your photo with the free tool and compare its explanation with what you see.

The 30-Second Check

Start with the details most likely to give you a clear question to investigate. Open the largest version you can find; a small thumbnail can hide a broken finger, a swapped letter, or a shadow that points the wrong way. Make one pass through these areas before deciding whether the image needs closer inspection:

  1. Hands: Follow each visible finger from knuckle to tip. Count only the fingers you can actually trace, then look at how they wrap around objects or disappear behind them.
  2. Text: Read signs, labels, clothing prints, and small captions letter by letter. Look for characters that change shape halfway through a word or words that make no sense in context.
  3. Light: Pick the strongest light source and compare shadows and shiny surfaces against it. A reflection should show the scene from the reflective surface's viewpoint, not a different arrangement of objects.

If those details look normal, the image may still be synthetic. If one looks wrong, first consider ordinary explanations such as motion blur, a wide-angle lens, compression, editing, or an object hidden outside the frame. The goal of this quick pass is to find a specific inconsistency you can verify, rather than to label the entire image from a vague impression.

The 9 Signs

Use the following signs as checks, not as a scorecard. Image generators vary, and real photographs can contain visual mistakes caused by cameras, software, or unusual scenes. A useful sign is one you can point to precisely: a joint that cannot connect, a word that cannot be read, or a reflection that does not match the room.

Hands and Fingers

Trace every finger through its joints, especially where a hand grips a cup, rests on another hand, or passes behind clothing. Check whether a fingertip belongs to a visible knuckle and whether the thumb sits on the correct side of the palm. Hands combine small shapes, flexible poses, and partial occlusion; an image generator can make each patch look plausible while losing the structure that connects them. A real photo can also hide fingers, so change your viewing angle mentally before calling an extra or missing digit.

Teeth and Eyes

Zoom in on a smile and compare tooth edges, gaps, and the line where teeth meet the lips; then compare gaze direction and catchlights in both eyes. Watch for teeth that merge into a smooth strip or eyes that seem to focus on different places without an obvious reason. These are small repeated structures whose parts must stay consistent across the face, and that consistency can break even when the portrait looks convincing at normal size. Reflections in eyes depend on the lighting setup, so they are useful only when you can compare them with the rest of the scene.

Text and Lettering

Read every word on signs, packaging, badges, and clothing rather than glancing at the overall shape of the letters. Check whether repeated letters keep the same form and whether a line of text follows the surface it is printed on. The image has to satisfy spelling, letter shapes, spacing, and perspective at once, which creates more opportunities for a mismatch than a plain patch of color. Blurred or partly hidden real text can also look strange, so give the image the benefit of that possibility.

Reflections and Shadows

Choose one mirror, window, glossy table, or shadow and compare it with the object that should produce it. Look for a missing object in a reflection, a face turned the wrong way, or shadows that imply incompatible light sources. Reflections and shadows are linked to the positions of objects and lights throughout a scene; a locally believable highlight may still fail that larger relationship. Multiple lamps, fill light, and edited composites can complicate a real photo, so judge the whole lighting setup before drawing a conclusion.

Background Objects

Inspect furniture legs, railings, shelves, utensils, and people at the edge of the frame, then follow their outlines through overlaps. A chair leg that fades into the floor or a railing that changes spacing halfway across is a concrete clue. Small background elements must keep their shapes and relationships even when they are partly hidden; generation can preserve the scene's general mood while letting those local connections drift. Depth blur and compression can erase similar details in a camera image, so check the sharpest available copy.

Depth of Field

Compare objects at roughly the same distance from the camera and ask whether their sharpness changes for a visible reason. An abrupt switch from crisp to smeared detail across one flat surface deserves a closer look. Real focus blur follows the optics and depth of the scene, while generated blur may imitate the look of a portrait without maintaining those relationships everywhere. Phone portrait modes also estimate depth and can blur the wrong edges, making this a weaker sign on its own.

Skin and Hair Texture

Look along the hairline, around ears, and where loose strands cross a face or garment. Notice whether strands suddenly merge, skin pores vanish in one patch, or fine texture changes at an unexplained boundary. These areas require tiny details to remain continuous across changing surfaces and lighting, so a polished overall portrait can still contain local transitions that do not connect. Makeup, retouching, compression, and shallow focus can create similar effects in genuine photographs.

Metadata and Watermarks

If you have the original file, inspect any available origin or edit information and check whether a credible source can explain where the picture came from. A supported origin record or detectable watermark can add evidence that appearance alone cannot provide because it records something about creation or editing. Such information may be absent from a screenshot, stripped during sharing, or changed by software; its absence does not establish that the picture came from a camera. Treat a visible logo or watermark as a claim to verify, since anyone can place a mark on an image.

The “Too Perfect” Tell

If every face is flawless, every surface is spotless, and the lighting feels uniformly cinematic, slow down and inspect a few ordinary details. That impression can arise because generation favors a coherent visual style while overlooking the messy variation of a particular real scene. It is the least specific sign here: studio lighting, makeup, beauty filters, and careful retouching can produce the same finish. When the image still leaves you unsure, upload an image for an AI check and read the result alongside the evidence you found.

What No Checklist Can Catch

A well-made AI image may have readable text, normal hands, and consistent light. A real photograph may contain an apparent extra finger because two hands overlap, strange lettering because a sign is out of focus, or impossible-looking shadows because the scene has more than one light source. The absence of visible errors therefore does not verify that a camera captured the scene, and the presence of one error does not settle its origin.

Copies make the problem harder. A screenshot can remove information attached to the original file, while cropping can hide the object that explains a shadow or reflection. An edited camera photo can also combine real and generated content, so a single yes-or-no label may oversimplify what happened. If the image matters, ask who published it, where an earlier version appears, and whether the original file or creator is available.

Use a detector as another signal, especially when your visual inspection finds nothing decisive. Its assessment can direct attention to details you missed, but it cannot replace source checking or prove a complex editing history. When the clues conflict, say that the origin is uncertain instead of forcing a confident answer.

Check a Photo in Seconds

After the visual pass, open the checker on the homepage and upload the image you want to examine. Read both the verdict and its plain-English explanation, then return to the original image to inspect the details that matter. A result is most useful when it gives you a testable reason to look again, such as inconsistent texture or a suspicious object boundary.

For the clearest comparison, use the original file when you have it. If you only have a social-media screenshot, keep that limitation in mind: it may be smaller, compressed, cropped, or missing origin information. Check the publisher and look for earlier appearances of the image if the claim attached to it is important. This process gives you a grounded answer even when the visual clues remain ambiguous.

FAQ

How to identify AI-generated images?

Start with text, hands, and lighting, then ask whether the background and reflections describe one coherent scene. If you have the original file, inspect any available origin information and search for earlier uses of the image. Treat each clue as evidence to weigh, not a final verdict.

How do I identify a picture using AI?

Upload the picture to an AI image detector and read the explanation alongside your own visual checks. A detector result is an assessment, so compare it with the picture's source and original file when the result is uncertain. A screenshot or heavily edited copy may leave less information to inspect.

Can ChatGPT identify an image?

ChatGPT can inspect an uploaded image and point out details worth checking, but its interpretation can be wrong. Ask it to describe specific visible inconsistencies rather than treating its answer as proof of origin. ChatGPT does not process the original file's metadata when analyzing an image.

Can Google detect AI images?

Google can provide image source context, and Gemini can check certain origin signals. Google's current watermark verification recognizes content created or edited by Google AI tools; a negative result does not rule out other AI systems. Compare that result with the image's source, visible details, and any available origin information, then get a photo verdict and explanation if you want a separate assessment.