# Can Google tell if an image was made with AI?

Source: https://creatisimo.net/topic/can-google-tell-if-an-image-was-made-with-ai/
Site: Creatisimo, category Content credentials and C2PA
Published: 2026-10-30. Updated: 2026-10-31. Replies: 8.

## Question by Ivy C., 2026-10-30

Can Google tell if an image was made with AI? I am a fact-check intern at a magazine, and my editor asked me to find out what Google actually does when you ask it about a suspicious image. I have seen "About this image" in search results, I have heard of SynthID, and I have read that Gemini can check files, but I cannot work out how these fit together.

If I right-click a picture in Chrome or upload it to Google, will it say "AI generated"? Does it work on images from other companies' tools, or only Google's? I need to explain this to the rest of the desk in a few sentences, and I do not want to overstate what any of it can do. A plain description of what each Google feature checks would help a lot.

## Reply 1 by Mei H., 2026-10-30

Google Lens will just tell you whether it is AI, I thought? I have seen people say it flags images instantly. I use it to find where my stock photos got copied, and it always gives me an answer.

## Reply 2 by Remy F., 2026-10-30

Lens finds visually similar images. Reading provenance is a different job. It reads a manifest that is embedded in the file, so a stripped copy shows nothing. I would not assume it labels anything on its own.

## Reply 3 by Oskar Lund (staff), 2026-10-30

Yes, partly: Google's SynthID check in Gemini, Search and Chrome detects images made with Google AI, and Search's About this image shows C2PA credentials when an image carries them. It cannot label every AI image.

Ivy, here is a desk-ready breakdown.

| Google feature | What it checks | Limit |
| --- | --- | --- |
| SynthID check in Gemini, Search and Chrome | An invisible watermark in images, video or audio made with Google AI | Gemini recognizes only Google AI content; "not detected" can still be another AI system |
| SynthID Detector portal | Watermarks from Google or partners named by DeepMind: OpenAI, NVIDIA, Kakao | Only content from tools that embed a SynthID watermark |
| About this image in Search | Whether an image was created or edited with AI tools, if it has C2PA metadata | Needs embedded C2PA data to show anything |

Google said in September 2024 that Search's "About this image" feature would let people see whether an image was created or edited with AI tools if it contains C2PA metadata, and that the feature is available in Google Images, Lens and Circle to Search. Google's own post says there is "no silver bullet solution for all content online."

Google DeepMind says SynthID adds an invisible watermark at the moment of creation, designed to survive cropping, filters and lossy compression. Its SynthID Detector can check whether content was made with AI from Google or partners including OpenAI, NVIDIA and Kakao.

Mei, Lens is not a general AI labeler. The label appears only when credentials or a watermark exist. A missing watermark means the image was not made with Google AI, and Gemini's help page adds that other AI systems could have made it.

For your desk: "Google can confirm its own watermark and read credentials when present. It cannot prove an image is real or fake." Sources: [How we're increasing transparency for gen AI content with the C2PA, Google blog, 2024](https://blog.google/technology/ai/google-gen-ai-content-transparency-c2pa/) and [Google SynthID, Google DeepMind](https://deepmind.google/models/synthid/). On survival of the watermark, see [the SynthID cropping thread](/topic/does-an-invisible-ai-watermark-like-synthid-survive-cropping/).

## Reply 4 by Nina W., 2026-10-30

Does a retouched photo of mine get flagged if I edit it in Photoshop? I am trying to figure out whether normal edits could get a real portrait labeled as AI.

## Reply 5 by Ezra M., 2026-10-30

My caution is the opposite one. People hear "Google can check" and treat a clean result as proof of authenticity. Nothing in what Oskar quoted says that. It is a check on one watermark and one metadata record, so silence is not clearance.

## Reply 6 by Oskar Lund (staff), 2026-10-30

@Ivy C. Context. It reports what the embedded record says about how the file was made. It does not rule on whether the scene is true, so quote it as "the credential states" rather than "Google says."

## Reply 7 by Ivy C., 2026-10-31

@Oskar Lund thank you, that table is exactly what I needed. Is the "About this image" result a verdict or just context?

## Reply 8 by Remy F., 2026-10-31

@Nina W. A credential from an editing app lists edits made with that app, and whether AI tools were used is recorded in the manifest. A normal retouch is not the same as generation, so read what the credential says before assuming anything.

---
Cite as: Creatisimo, "Can Google tell if an image was made with AI?", https://creatisimo.net/topic/can-google-tell-if-an-image-was-made-with-ai/, accessed 2026-10-31.