OpenAI is adding Google SynthID watermarks to ChatGPT, Codex, and API images, combining hidden signals and C2PA metadata for origin checks in Search and Chrome.
During Google I/O 2026, OpenAI has announced it is bringing Google’s SynthID to AI-generated images while previewing a tool that verifies OpenAI images. OpenAI gains a new way to test whether an image came from its own systems after that file leaves ChatGPT or the API.
At launch, SynthID integration will start with images created with ChatGPT, Codex, and the API. Because the rollout starts with those products, the verifier works as a product-origin check rather than a scanner for every AI image online.
C2PA conformance means a file can carry content credentials that describe how media was made or edited. Because SynthID adds an imperceptible marker, Google’s earlier C2PA authenticity work and its public detector give the checking side a consumer-facing precedent.
OpenAI’s verifier is designed to check both metadata and watermark signals in supported images. Because a metadata label can break or disappear as a file moves between services, a hidden watermark gives the verifier another signal to inspect.
OpenAI now combines C2PA metadata with SynthID in AI-generated images, so the verifier can compare a visible record with a hidden signal.
Using both layers gives OpenAI a stronger backup when metadata tagging was pretty easy to defeat, though it does not make the verifier a universal detector for every model or edit chain.
Google is pushing the checking side closer to ordinary web use so people can question images in Search or Chrome rather than inside a developer workflow.
“We want more people to have easy access to these tools, so we’re expanding both Content Credentials and SynthID verification to Search and Chrome.”
OpenAI’s first verifier remains limited to images generated by OpenAI products. Because the first release stays inside OpenAI-origin media, it remains useful for that narrower category while leaving other models, platforms, and editing pipelines outside the initial scope.
After its image generation API release, its watermark experiments, and Google’s public detector, OpenAI still does not have a universal checker that spans every model and editing chain.
Google has expanded verification tools across Google products, including Search and Chrome. Chrome and Search placement builds on Google’s 2024 C2PA groundwork and gives OpenAI’s watermarking move more practical value than a label buried inside one app.
Pixel 10 camera support gives Content Credentials a native capture role. Incoming Pixel video support would move provenance checks from still images into more everyday recording workflows, while Google’s AI Content Detection API on its Gemini Enterprise Agent Platform extends the same authenticity push into cloud tooling.
Google puts SynthID’s scale at watermarked more than 100 billion images and videos, plus 60,000 years of audio. OpenAI’s support plugs into a provenance system Google already treats as infrastructure, not just a marker attached to a single product.
Google is expected to integrate SynthID with Circle to Search, Google Lens, and AI Mode, alongside Chrome. Wider placement would move authenticity checks closer to the moment users encounter a disputed image.
There is currently not a public API for SynthID, with Google framing broad scan access as a potential attack vector. Keeping the API closed widens visibility for users while limiting tools that could help attackers tune around detection.
Google had already moved SynthID into edited-image workflows through Google Photos in February 2025 and added a public SynthID detector in May 2025.
OpenAI’s earlier developer rollout and watermark testing phase give the current verifier a clearer product lineage.
Google blog authors Laurie Richardson and Pushmeet Kohli described OpenAI, Kakao, and ElevenLabs as part of a broader push to bring SynthID to AI-generated content. For users, the practical change is narrower: OpenAI’s first verifier checks origin for images from its own products, not all AI media circulating across the web.
OpenAI could expand the verification tool over time, but the 2026 first release remains limited to three starting sources. Until OpenAI expands the verifier beyond ChatGPT, Codex, and API images, users can treat it as an OpenAI-origin check rather than a universal AI-media scanner.
Leave a Reply