Gemini verifies AI-generated images from Google

Google now lets you check in Gemini whether an image was created or edited with the company's artificial intelligence tools. Just upload it to the app and ask: "Was it created with Google AI?" or "Was this image generated with AI?"

Gemini will look for an invisible digital mark called SynthID. This technology embeds imperceptible signals in content generated by Google's models, allowing it to be identified later without changing its appearance.

Gemini's response does not prove whether an image was created with any AI system. It indicates whether it finds Google's mark and adds context about the content you're reviewing.

How verification works

The process is simple:

• Upload the image to the Gemini app.
• Ask whether it was created or edited with Google AI.
• Gemini checks whether it contains a SynthID mark and explains the result.

The tool can be useful when you receive a viral image, a screenshot from social media, or an alleged photograph that looks real but whose origin you're not sure about. That said, if Gemini doesn't find the mark, it does not automatically mean the image is authentic: it may have been created with another model or lost the signal during a modification.

Google introduced SynthID in 2023. Since then, the company says more than 20 billion pieces of AI-generated content have been marked with the technology. Until now, it had also tested a standalone verification portal with journalists and media professionals.

Verification will come to video and audio

Google plans to expand SynthID to other formats, including video and audio, and bring these features to more products, such as Google Search. The company is also working with the Coalition for Content Provenance and Authenticity, known as C2PA, which promotes standards for recording the origin and history of digital files.

As part of that work, images created this week with Nano Banana Pro, also known as Gemini 3 Pro Image, in Gemini, Vertex AI, Google Ads, and Flow will include C2PA metadata. This data can provide additional information about how an image was created.

Later, Google wants its systems to be able to read C2PA credentials from content created outside its own ecosystem. In practice, that would make it possible to check the origin of images produced with other companies' tools, as long as they include that data.

The key point is that verifying an image is not the same as certifying that it is true. For now, Gemini can help you determine whether an image carries signals from Google and, over time, it could offer more clues about its origin. Its usefulness will depend on more platforms adopting provenance standards and those marks surviving when content is copied, edited, or shared.