Image Tools
AI image detector
Upload an image to check whether it's AI-generated or a real photo. The tool inspects metadata, generator signatures, content credentials, and image dimensions — all locally in your browser — and shows you exactly which signals it found.
Drop an image to analyse
or click to browse — JPG and PNG work best
What the tool looks for
Real photos carry a fingerprint from the device that took them: the camera make and model, exposure settings, and often a timestamp, all stored as EXIF metadata. AI image generators almost never write this data. Instead, they sometimes leave their own marks — a software tag naming the generator, a C2PA content-credentials manifest, or, in the case of Stable Diffusion and ComfyUI, the full prompt and generation parameters tucked into the PNG's text chunks.
This detector scans for all of those signals and weighs them. Camera EXIF pushes the verdict toward “real photo,” while generator signatures, embedded prompts, and default square dimensions (like 512×512 or 1024×1024) push it toward “AI-generated.” Every signal it finds is listed with a plain-English explanation, so you can judge the evidence yourself.
Why no detector is perfect
Honesty matters here. Metadata-based detection is reliable only when the metadata survives — and most of the time it doesn't. The moment an image is screenshotted, uploaded to social media, or re-saved in an editor, its original metadata is usually stripped. At that point even expensive, pixel-analysis-based detectors struggle, frequently getting 20–40% of cases wrong in independent tests.
Metadata can also be faked: someone can add fake camera EXIF to an AI image, or strip the generator tag from one. So use this tool the way a journalist uses a source — as one piece of evidence among several, not as the final word. When a result is “inconclusive,” that's the tool being honest that the file simply doesn't contain enough to say.
Frequently asked questions
- How does this AI image detector work?
- It reads the file locally and looks for signals that correlate with AI generation: missing camera EXIF data, AI generator software tags (Midjourney, DALL·E, Stable Diffusion, Firefly), C2PA content credentials, Stable Diffusion generation parameters embedded in PNG text, and telltale square dimensions like 1024×1024. Each signal is shown with an explanation.
- Is the result 100% accurate?
- No, and no detector is. This tool is an indicative heuristic, not proof. It's strong when metadata is intact, but screenshots, social-media re-saves, and metadata stripping erase the clues — and metadata can be deliberately faked. Treat the result as evidence to weigh, not a final verdict.
- Why does it say 'inconclusive' for a screenshot?
- Screenshots and re-exported images lose their original metadata, so there's little for the tool to analyse. That's why platforms like social media make AI detection so hard — the very clues a detector needs are gone by the time you download the image.
- Does my image get uploaded?
- No. Everything runs in your browser. The image is read into memory locally, analysed, and never sent to any server. You can disconnect from the internet and it still works.
- What are C2PA content credentials?
- C2PA (Content Credentials) is an industry standard for attaching tamper-evident provenance to media. Adobe Firefly, OpenAI, and others embed it to mark AI-generated or AI-edited images. When present, it's a meaningful signal — but its absence doesn't prove an image is real.
- Can it detect AI-edited (not fully generated) photos?
- Sometimes. AI editing tools may leave software tags or C2PA manifests that this tool flags. But subtle edits to an otherwise real photo often leave no metadata trace, so they can pass as real.