How to tell if an image was made with Stable Diffusion

What Stable Diffusion writes into the file

Stable Diffusion is an open model. Which traces end up in the file depends mainly on the program, and Stability AI's hosted services behave differently from programs on your own computer.

C2PA (Content Credentials) is a signed provenance record in the file, IPTC Digital Source Type a metadata field on how the image was made, for example trainedAlgorithmicMedia for fully AI-generated images. An invisible watermark sits in the pixels, and only the vendor can read it.

How to check a Stable Diffusion image

1. Drag the original file from Stable Diffusion into the box above. It is read in your browser, not uploaded. Screenshots and copies from stores or social networks usually no longer carry the data.
2. If a PNG file contains one of the fields from the table, the checker names the program and shows prompt, negative prompt, seed and model under “Prompt and settings”.
3. The prompt view currently only reads PNG fields, not the EXIF field UserComment that AUTOMATIC1111 and Forge use for JPEG and WebP.
4. If the image comes from the Stability AI API, the checker shows the C2PA record under “Content Credentials (C2PA)”.

More detail in the Prompt reader and AI image checker and Content Credentials viewer.

The checker is not an AI detector. It only shows what the file says. An empty result does not prove a real photo, not least because providers of AI systems placed on the market before 2 August 2026 have until 2 December 2026 to add machine-readable marking (Article 111(4) AI Act).

What survives uploads and edits

The prompt fields are notes kept by the program, not provenance records, and only exist in the file the program saved. A screenshot does not carry them, and any program that saves the file again may leave them out.

Saving them can be switched off in these programs, and Fooocus has it off by default. The checker cannot read an invisible watermark, if there is one.

Labelling duty in your store under the EU AI Act

If you use Stable Diffusion to create or edit images for your own store, you are a deployer, even when an agency or an employee works on your instructions (Article 3(4) AI Act, Commission guidelines para. 12, 14). If the image is a deep fake, meaning realistic and falsely appearing authentic, you have had to label it visibly since 2 August 2026 (Article 50(4) AI Act). The labelling check tells you whether that applies, and the guide to the labelling duty explains the rules.

The traces in the table do not replace that label, because people cannot see them straight away (guidelines para. 117). Still, keep them in the file, as the guidelines recommend. They document where the image came from.

The stored prompt shows what was meant to be generated, such as a person or a whole scene. That helps you decide whether the image is a deep fake, and it documents your decision.

How Stable Diffusion labels its own output

Stability AI seals the C2PA record in its API and hosted services with its own certificate, and says no provenance is applied when its openly released models generate images. According to Stability AI, other watermarking methods degraded image quality.

The local programs store the prompt and settings mainly so that an image can be recreated later. As far as our research shows, they add no visible mark by default.