Is This Image AI-Generated — or Fake? Check It in Seconds

How to tell if an image is AI-generated

Hands, fingers and limbs. Count fingers and look at joints, fingernails and held objects. Merged or inconsistent anatomy is a reason to look closer, but motion blur, perspective and unusual poses can also look strange in real photographs. No single detail is a reliable verdict.

Text inside the image. Check signs, labels, book spines and screens. Look for inconsistent lettering or words that change across an image. Small or compressed text can be hard to read in real photos too, and generated text can be convincing.

Reflections, shadows and light sources. Compare shadows and reflections with the scene. An apparent mismatch deserves investigation, but multiple light sources, mirrors and glass can produce unexpected effects in a genuine photo.

Repeated or blended detail. Inspect hair, leaves, fabric, background windows and distant faces. Repetition or merged objects may prompt a closer review. Patterns alone do not establish how the image was made.

Backgrounds and edges. Look at where the subject meets the background. Halos, unusual blur or cut edges can arise from editing, camera processing or compression as well as generation.

Small hardware. Check whether earrings, glasses, buttons, straps and necklaces connect in a plausible way. Compare with the full scene before drawing a conclusion from one detail.

What does not establish AI generation. Noise reduction, shallow depth of field, warm colour grading, HDR skies, smooth skin and lens flare all occur in real photography. On their own, they are not evidence of AI generation.

Review the whole image. Compare several clues, find the original source and consider ordinary explanations. These are manual review suggestions, not a list of features our classifier measures or highlights. The automated check returns a classification and model confidence.

Is this image fake — or was it AI-generated?

These are different questions. A fully AI-generated image is synthesized rather than captured as that scene by a camera. A fake or misleading image may instead be a real photograph that was altered, or an unchanged photograph presented with a false claim.

A face swap, changed document amount, removed object, old photo presented as new or stock photo presented as a personal snap all require context. A classifier cannot verify the claim attached to a picture.

If your question is “does this look AI-generated?” Upload the file for an AI / human classification, then review its limitations. A human-made prediction does not establish that a camera captured the claimed event.

If your question is “has this photo been altered?” Read the fake image detector guide for editing and context checks. It uses the same classifier; it does not add a manipulation heatmap or specialized face analysis.

If your question is about a painting or illustration. Use the AI art detector page for artwork-specific review guidance and advice on preserving evidence of your process.

What the report shows

1. The classification. An AI-generated or human-made prediction with model confidence. This score is not an independently verified probability that the image is authentic. False positives and false negatives are possible.

2. A record of the check. The report includes the image, check time and available model output. You can export JSON or print it. It does not contain regional explanations, four forensic verdicts or a C2PA verification result.

3. What to do next. Compare the original file, look for earlier copies with reverse image search and ask the sender for context. Keep a copy of the report before its 24-hour expiry if you need a dated record.

A score is the beginning of a review

1. Read the result carefully. Imagefake returns an AI or human-made classification and its model confidence. The score is not an independently verified probability of authenticity.

2. Find the original source. An original camera file may retain metadata and more detail than a screenshot or forwarded copy. Metadata can be missing or changed, so neither its presence nor absence proves authenticity. Try reverse image search for earlier versions.

3. Check the context. Who sent it, where it was posted, when, and what claim is attached to it all matter. A detector cannot establish why someone sent you a picture or whether its caption is true.