ChatGPT image detector: was this made with ChatGPT / DALL·E?
Drop an image to look for the labels OpenAI adds to ChatGPT / DALL·E images. Also detects other AI tools. Nothing is uploaded.
How ChatGPT / DALL·E images are labeled
OpenAI attaches C2PA Content Credentials to images created in ChatGPT and through its image API. When the original file is checked, the manifest names OpenAI as the issuer and declares the image as AI-generated.
| Made by | OpenAI |
|---|---|
| Also known as | DALL·E 3, GPT-4o image generation, GPT-Image, Sora images |
| Content Credentials (C2PA) | Yes |
| IPTC “AI-generated” label | Sometimes |
| Prompt / settings saved in file | No |
| Invisible watermark | Not confirmed |
Last reviewed November 2026. AI companies change their labeling — we update this page regularly.
What this checker looks for
We look for the C2PA manifest, the OpenAI / ChatGPT names inside it, and the IPTC “trainedAlgorithmicMedia” label.
Other ways to check ChatGPT / DALL·E images
- Screenshots and re-uploads to social media remove Content Credentials — ask for the original download if you can.
- Look for typical ChatGPT image traits: very clean lighting, slightly warm colour cast, and garbled small text in backgrounds.
5 ways to tell if an image is AI-generated
- Check the metadata (this tool). ChatGPT, Adobe Firefly, Microsoft and others attach Content Credentials or IPTC labels. Stable Diffusion apps store the prompt in the PNG.
- Check for invisible watermarks. Google's SynthID survives screenshots; the Gemini app can check whether an image was made with Google AI.
- Reverse image search. Google Lens or TinEye can reveal where an image first appeared — often an AI art gallery or a post saying it's AI.
- Look closely. Hands and fingers, jewellery, teeth, text and signs, reflections, and repeated background patterns are still common giveaways.
- Check the context. Who posted it first? Do reputable sources show the same scene from other angles?
Honest limits: no tool can prove an image is real. A missing AI label only means there's no label. Treat a positive label as strong evidence and a negative result as “unknown”.
Frequently asked questions
Can you tell if an image was made with ChatGPT?
If you have the original file downloaded from ChatGPT, usually yes — it carries Content Credentials naming OpenAI. If it was screenshotted or shared through social media, the label is often gone and no metadata checker can confirm it.
Is this chatgpt image detector free?
Yes — free, no sign-up, and the image is checked inside your browser without being uploaded.
What if no AI label is found?
That doesn't prove the image is real. Labels are removed by screenshots, re-saving and most social networks. Try the other checks on this page.