OpenAI is stepping up its fight against misleading AI-generated content by introducing two new safeguards aimed at identifying fake images online. The company revealed the updates on May 19, including a collaboration with Google and the rollout of a public image verification system.
As part of the partnership, OpenAI will begin embedding Google’s invisible SynthID watermark into every image produced through its AI platforms. Alongside this, the company has launched a verification feature that allows anyone to check whether an image was created using OpenAI’s tools.
The verification system works by scanning uploaded images:
The verification system works by scanning uploaded images for two indicators. The first is SynthID, Google’s hidden watermarking technology, while the second is C2PA metadata, an industry standard that labels AI-generated content through embedded file information.
At the moment, the checker is limited to visuals created with ChatGPT, Codex, and OpenAI’s API services. However, OpenAI says it plans to broaden support for content generated by additional AI systems in the future.
“At launch, the tool is limited to content generated by OpenAI. In the upcoming months, we aim to support cross-industry efforts to make verification possible across platforms. Over time, we also expect to support more types of content that people may encounter online, OpenAI said in a blog post.
The announcement comes at a time when AI image generators are becoming more advanced and easier for the public to access. As synthetic visuals continue to improve in quality, distinguishing between real and fabricated images has become increasingly difficult.
Although OpenAI’s latest measures could help reduce the spread of manipulated media, experts note that many questionable AI tools operating outside major platforms may still contribute to online misinformation.
Here’s how users can try the new verification tool:
Upload a single image file to the checker
Supported formats include PNG, JPG, and WEBP
The system will analyze the image for C2PA metadata or the hidden SynthID watermark
Results will indicate whether the image contains recognized AI-generated signals
OpenAI also recommends uploading clean, uncropped files whenever possible. Images containing multiple pictures or heavily edited screenshots may reduce the accuracy of the detection process.
OpenAI’s verification tool operates primarily as an open-web portal designed to catch content specifically generated within its own ecosystem (like DALL-E or ChatGPT) using a combination of cryptographic C2PA metadata and embedded watermarks. Google’s approach, powered by its native SynthID technology, is more deeply integrated into its ecosystem—such as through Gemini.