Verify OpenAI-generated content

Provenance helps people understand where online content came from and whether AI was involved in creating or editing it. This gives people more information when deciding what to trust, share, or report online.

It can help reduce confusion, deception, and cases where the origin of AI-generated content is unclear. It also gives journalists, platforms, creators, researchers, and the public more context about the media they encounter online.

We believe clearer information about how content was created is an important part of building a more trustworthy online environment.

The tool relies on two types of signals: C2PA Content Credentials and SynthID.

Content Credentials⁠(opens in a new window) are based on an open standard that attaches provenance metadata to digital media. For supported OpenAI-generated content, they can indicate that a file was created using tools such as ChatGPT, Codex, or the API. They are a strong signal when present, but because they rely on metadata, they can be removed.

SynthID⁠(opens in a new window) is an invisible watermarking technology that embeds a signal directly into generated media. Unlike metadata, the signal is part of the content itself and is designed to stand up to modifications like cropping, adding filters, or lossy compression. For more technical details, see the SynthID research paper⁠(opens in a new window).

A detected signal indicates that the content likely originated from OpenAI tools. Detected signals are reliable, and false positives are rare.

It does not confirm how the content has been used or modified after it was created.

The tool detects content that was generated by our tools, which includes cases where a user asked for a modification of an uploaded file, and the model generated a new file that reflects the changes.

No, a detected signal indicates that the content originated from OpenAI tools, but it does not determine whether the content is accurate or presented in the correct context.

For the most reliable results, upload a single file at a time.

For images, avoid cropping the image or converting it to another file format (for example, PDF). Upload only one image at a time.

For audio, upload clips between 10 and 60 seconds long for the best results.

If no signal is detected, it means the tool did not find supported signals in the uploaded file.

The content could still have been generated by OpenAI if its metadata was stripped or tampered with, its watermark was degraded, it came from a legacy generation model, or it was created before provenance signals were available. Content could also still be AI-generated by another company’s model, which the tool currently does not detect.

If the tool isn’t working, try again in a moment—some errors (like rate limits or temporary issues) can occur.

If the problem persists, visit our Help Center⁠(opens in a new window) for support.

The uploaded file is processed to check for supported provenance signals associated with OpenAI-generated content. Uploaded files are not stored unless legally required and are not used to train our models.

By using this tool, you agree to the OpenAI Terms of Use.

Yes. Developers can use the OpenAI API to verify content programmatically. See the API documentation⁠(opens in a new window) to learn how to get started.