Which AI Image Generators Use SynthID?

SynthID began as a Google technology, and Google's image and video models are still where you're most certain to find it. But in 2026 the picture is shifting: other major providers have started adopting SynthID-style watermarking, while a lot of open-source models don't embed it at all. So the accurate answer is "Google for sure, a growing list beyond that, and many open models not at all" — and it's changing fast enough that you shouldn't assume.

Here's the current lay of the land.

Google's tools: where it began

SynthID is a Google DeepMind technology, and it's applied across Google's generative models — Imagen for images, Gemini's image generation, and Veo for video. If an image came from one of those, you should assume it carries the watermark. Google has also begun surfacing AI-content detection in Chrome and Search, which makes the watermark more consequential the more places learn to read it.

Adoption is spreading in 2026

The big shift this year is that SynthID stopped being Google-only. Notably, OpenAI began adding support for SynthID watermarks in 2026 — a strong signal the standard is moving toward becoming an industry norm rather than one company's project. Expect the list of tools that embed it to keep growing.

The practical consequence: a watermark that was easy to avoid a year ago by simply "not using Google" is becoming harder to dodge by brand alone. What matters is the specific model under a given tool.

Tools that generally don't embed it

On the other side, many open-source image models don't apply SynthID. A number of AI photo services are built on these open models, which is why their output can ship without the pixel watermark. This isn't a blanket guarantee for every tool — implementations vary — but it's the category where clean output is common.

Because it varies, transparency matters. A tool that's clear about what it runs on lets you reason about whether SynthID is in play; a black box doesn't.

Why you can't assume

Two reasons "it probably doesn't have it" is a weak plan:

1. Adoption is expanding, so today's safe assumption may be wrong next quarter.
2. You can't easily verify — there's no public detector to read the watermark, so you're reasoning from the source, not confirming from the file. See how to check if an image has SynthID.

When you can't verify after the fact, the source you choose becomes the decision.

What to do for dating photos

If a watermark-free file matters to you, choose a generator that doesn't add one in the first place — because removing SynthID later isn't realistic. That's how our AI dating photos are built: open models that don't embed SynthID, plus a clean re-encode that drops metadata too. For picking a tool more broadly, see our take on the best AI dating photo generator and the full SynthID guide.

Frequently Asked Questions

Does Google use SynthID on all its AI images? Google applies SynthID across its generative models (Imagen, Gemini image generation, Veo). If an image came from a Google AI tool, assume it's watermarked.

Do other AI companies use SynthID now? Adoption is spreading. In 2026, OpenAI began adding support for SynthID watermarks, and the standard appears to be moving toward broader industry use.

Do open-source models add SynthID? Many don't, which is why some AI photo tools ship watermark-free files. It's not universal, though — implementations vary, so transparency about the underlying model matters.

How do I avoid SynthID entirely? Use a real photo or a generator built on open models that don't embed it. Since you can't reliably remove it, choosing the source is how you control it.