On May 19, Google announced at Google I/O 2026 that it would extend SynthID content verification capabilities to Google Search, Chrome, and the Gemini app. Moving forward, users can ask Lens, AI Mode, or Circle to Search whether a given image was generated by AI; Gemini will then provide a detailed provenance report based on embedded metadata such as device information and editing history. Chrome is set to follow suit within the coming weeks. Meanwhile, OpenAI simultaneously revealed plans to layer Google DeepMind’s invisible SynthID watermark atop its existing C2PA content credentials framework. Initially, this will apply to images produced by ChatGPT, Codex, and the OpenAI API, while OpenAI also previewed Verify — a public tool enabling anyone to verify whether an image originates from an OpenAI model. According to Google, SynthID has been utilized over 50 million times worldwide since Gemini gained image, video, and audio detection features.
Technically speaking, SynthID embeds imperceptible signals directly into pixel or audio data, rendering them resilient against cropping, compression, and screenshotting; C2PA, conversely, utilizes cryptographic signatures to record creator identity and editing history at the file metadata level — together they form a complementary system which OpenAI describes as a “two-layered defense mechanism.” New partners joining the SynthID ecosystem this year include OpenAI, Kakao, and ElevenLabs; Nvidia joined last year. On the hardware front, Google now supports C2PA tagging for videos captured on Pixel 8 series devices and newer models (previously limited to photos taken on Pixel 10), while collaborating with Meta to automatically label authentic content originating from Pixel devices on Instagram as non-AI-generated. For enterprises, Google Cloud has launched a dedicated AI content detection API, with initial partners comprising Shutterstock, Avid, and Canva. Both Google and OpenAI emphasize that the overarching aim of these systems is to “raise the cost of misuse” rather than completely eradicate AI-generated forgeries; under certain deliberate circumvention efforts, limitations inevitably remain.