Gemini Watermark: What It Is, Who Gets It and SynthID

The Gemini watermark is a small visible sparkle that Google adds to the corner of some images made in the Gemini app, and it sits on top of a second, invisible mark called SynthID that is built into every image Google’s AI tools generate. Reporting from late 2025 said the visible sparkle appears for free and Google AI Pro users but not for Ultra subscribers, and that Google’s own tools can check an image for the hidden mark. This guide explains both marks, who sees which, how to check an image and how to label AI pictures honestly. Google changes these rules, so confirm current details in the Gemini app.

Key takeaways

• There are two different marks: a visible sparkle in the corner and an invisible SynthID signal woven into the pixels.
• The visible sparkle was reported to appear for free and Pro users and not for Ultra users. Settings can change, so check your own output.
• SynthID is applied to images from Google’s AI tools whatever your plan, and it is designed to survive common edits such as resizing and compression.
• No detector is perfect. A missing mark does not prove a picture is real, and a found mark does not tell you who created it.
• This article does not explain how to remove either mark. Removing disclosure from AI images can mislead people and may break platform rules or local law.

What is the Gemini watermark?

The Gemini watermark is the visible four-point sparkle logo that Gemini places near the bottom corner of many generated images. It works like a small label that says “made with Google AI”. It is a visual cue for people, so it is easy to crop out, which is why Google also uses an invisible signal that is much harder to strip away.

Feature	Visible sparkle	SynthID
Can you see it?	Yes, in a corner of the image	No, it is embedded in the pixels
Who gets it?	Reported for free and Pro; not Ultra	Images from Google’s AI tools
Easy to crop out?	Yes	No, it is built to survive many edits
Purpose	A quick signal for viewers	A machine-readable check

Who gets the visible Gemini watermark?

TechRadar reported in November 2025 that Google planned to keep the visible sparkle on images from free and Google AI Pro users and to remove it for Ultra subscribers and some developer tools. The same report said the invisible mark would stay for everyone. Plan features and defaults change often, so treat this as a snapshot and look at an image you have just generated rather than relying on memory.

• Free accounts: expect the sparkle on generated images.
• Google AI Pro: reported to show the sparkle as well.
• Google AI Ultra: reported to be without the visible sparkle.
• Developer tools: behaviour can differ, so read the current documentation for your tool.

Our how we test page explains how we verify claims like these, and the disclaimer covers the limits of what we can promise when products change.

What is SynthID?

SynthID is a technology from Google DeepMind that embeds an imperceptible signal in AI-generated images, audio, video and text. For images, the signal is spread across the pixels in a way a person cannot see but a detector can read. Google designs it to hold up through common changes such as compression, resizing, filters and screenshots, though no method is guaranteed against every edit.

Because the mark is invisible, it does not change how the picture looks. Its job is to let a tool, and eventually platforms and publishers, work out whether an image came from a Google model.

How to check if an image was made with Google AI

1. Open the Gemini app or the web version while signed in.
2. Upload the image you want to check.
3. Ask a direct question, for example “Was this image created or edited with Google AI?”
4. Read the answer carefully. A positive result means a Google AI signal was found. A negative result only means none was detected.
5. Look for other clues, such as the visible sparkle, odd text, warped hands or inconsistent shadows.

Google has said it is expanding these checks, and the exact features may differ by country and app version. If you rely on the answer for something important, such as a news story, ask for the original file and the person who made it.

What the checks can and cannot tell you

Result	What it suggests	What it does not prove
Google AI signal found	A Google model likely generated or edited the image	Who made it or why
No signal found	Not detected by this check	That the image is real; other AI tools may not use SynthID
Visible sparkle present	Likely made in the Gemini app on a plan that shows it	That it was not edited afterwards
Sparkle missing	Could be Ultra, a different tool, a crop or a real photo	That the image is authentic

Should you remove the Gemini watermark?

We do not cover removal methods. A watermark exists so viewers can tell an AI image from a camera photo. Taking it off and presenting the picture as real can mislead people, and it may breach the terms of the service you used, a platform’s policy on synthetic media or laws in your area. Many sites now ask creators to label AI-generated content.

If the sparkle bothers you on a personal project, there are honest options:

• Frame the image so the corner has space, then keep the label visible.
• Add a caption such as “Created with AI” next to the picture.
• Consider whether a paid plan that does not show the sparkle fits your use, and still disclose AI use where it matters.
• Use real photographs when authenticity is the point, for example in news, reviews or evidence.

Labelling AI images honestly

• Say “AI-generated” or “created with Gemini” in the caption or alt text.
• Do not present an AI picture as a photograph of a real event.
• Do not create or share realistic images of real people without their permission.
• Keep the original prompt and file so you can show how the image was made.
• Follow the rules of each platform where you post.

If you make festive or portrait images, our guides to Gemini prompts and the Nano Banana prompt guide include notes on consent and labelling. You can find more in the Gemini and Nano Banana category.

What to do if you find a mark on an image you did not make

If you receive a picture that carries the sparkle, or a check flags a Google AI signal, treat it as a prompt to ask questions, not as a verdict. Ask the sender where the image came from and whether they edited it. Look at the context: a caption claiming a real event deserves more care than a clearly artistic piece. If the image is being used to mislead, such as a fake news photo or a false product review, report it on the platform where you saw it.

If you publish images yourself, a short note such as “Image created with Google Gemini” is enough for most readers. It costs nothing and builds trust, especially on sites that care about accuracy.

Common misunderstandings

• “No sparkle means it is a real photo.” Not necessarily. The sparkle can be cropped, absent on some plans, or the image can come from another tool.
• “SynthID shows who made the image.” It points to the model, not the person.
• “Screenshots wipe the mark.” Google designs SynthID to survive many such changes, but strong edits can weaken any signal.
• “All AI companies use SynthID.” No. It is Google’s system, and other companies use their own labels or none.

Frequently asked questions

Final thoughts

The visible sparkle is a courtesy label and SynthID is the sturdier, hidden layer behind it. Use the checker when you need a quick signal, remember its limits and label your own AI images clearly. Honest disclosure protects your readers and your reputation.

go

Written by Google Edits Team

The Google Edits team tests Gemini photo editing, Nano Banana and NotebookLM by hand and dates every guide. Independent, not affiliated with Google. How we test.

Independent publication. Not affiliated with or endorsed by Google. Limits and prices change; verify in your own account.