Images generated by ChatGPT will now feature Google’s SynthID watermarks.
OpenAI has announced a series of updates today, designed to significantly enhance the identification of content generated by its AI models online. Alongside reinforcing its commitment to embedding C2PA content credentials — widely recognized as the leading standard for verifying the provenance and edits of image, video, and audio content — OpenAI will now also integrate Google’s SynthID watermarks, implementing a “multi-layered approach” to AI labeling tools.
“These two systems reinforce each other. C2PA helps content carry detailed context; SynthID helps preserve a signal when metadata does not survive,” OpenAI stated in its announcement. The company further explained, “Watermarking can be more durable through transformations like screenshots, while metadata can provide more information than a watermark alone. Together, they make provenance more resilient than either layer would be on its own.”
Initially, SynthID watermarking will be applied to images generated by ChatGPT, Codex, or the OpenAI API. The emphasis on metadata preservation is particularly notable, especially as SynthID has anecdotally demonstrated greater reliability for fact-checkers and media agencies in verifying deepfake images online compared to C2PA. This expanded, dual-layered approach is expected to reduce the likelihood of OpenAI-generated content circumventing verification systems, thereby making deepfakes easier to distinguish and assisting online platforms in labeling AI-generated or manipulated content for their users.
As part of this comprehensive expansion, OpenAI is also previewing a public verification portal, which will allow users to determine if images contain AI metadata or watermarks. When an image is uploaded, the portal will examine C2PA and SynthID provenance signals to flag whether it was generated with ChatGPT, the OpenAI API, or Codex. While this service is initially limited to images generated by OpenAI, the company has stated its intention to support other verification systems in the coming months and eventually broaden its scope to cover more types of online content.
OpenAI acknowledges the inherent limitations of such technologies, stating, “No detection method is foolproof, so we take a cautious approach in cases when detection fails.” The company clarifies that if no metadata or watermark is detected, the tool will not make a definitive conclusion about whether the image was generated with OpenAI tools, as provenance signals can, in some instances, be stripped.
OpenAI has also joined the C2PA Conformance Program. According to the program’s description, this initiative “provides assurance that products adhere to the Content Credentials specification, and fulfill a set of security requirements to ensure they are producing and validating C2PA data correctly.” It is important to note, however, that while OpenAI has been embedding C2PA data into its image and video content for some time, this system has previously offered limited assistance in reliably identifying OpenAI’s deepfake content in wider circulation. This is primarily due to the ease with which metadata can be removed once content leaves its original platform, with some platforms even inadvertently stripping it during the upload process.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.