Google is embedding an AI image verification feature directly into its Gemini app. Users can upload a picture and ask whether it was generated or edited with Google’s models. The service scans for SynthID, an imperceptible watermark introduced in 2023, and returns a reasoned answer that adds provenance context to the visual content, and works entirely on‑device, preserving user privacy.

The rollout follows a year of watermarking experiments that have already tagged more than 20 billion AI‑produced assets. Google is also extending the same verification logic to video and audio formats and embedding C2PA metadata in images generated by Gemini 3 Pro, Vertex AI and Google Ads. These steps also ready the ecosystem for regulation. Partnerships with the Coalition for Content Provenance and Authenticity aim to broaden industry‑wide transparency.

By surfacing the watermark check inside a conversational interface, Gemini gives developers and everyday users a lightweight tool to assess authenticity without leaving the app. Early testers, including journalists, report that the detector’s reasoning clarifies ambiguous images, reducing the risk of misattribution. The feature now lives across Search, YouTube and Photos, cementing Google’s push for media. It demonstrates a path toward verifiable AI outputs.