Gemini watermark by tier: Free, Pro, Ultra, and API

The visible corner mark is a tier decision, not a universal Gemini behavior. Free and Google AI Pro image outputs carry a visible corner mark plus SynthID. Google AI Ultra and the Google AI Studio and Gemini API surfaces carry SynthID only, with no visible mark added. That split, stated in Google's own Nano Banana Pro announcement, is the entire answer to "why does this Gemini image have a watermark and that one doesn't."

It also means the useful move sometimes isn't a removal tool at all. If the image came from Ultra, AI Studio, or the API, there is no visible mark on it to begin with, and running any eraser against it will correctly find nothing to do.

Which Gemini tiers show a visible watermark?

Free and Google AI Pro outputs carry both a visible corner mark and SynthID. Google AI Ultra and the AI Studio and API surfaces carry SynthID only, with no visible mark. That is the whole rule; nothing else about a prompt, image size, or output resolution changes which side of that line an image lands on.

A few things follow from that table. First, the split tracks the generation surface, not the image's subject, aspect ratio, or output size. A square Free-tier render and a panoramic Ultra-tier render both follow the same rule for their own tier. Second, "Pro" here means the Google AI Pro subscription used inside the Gemini app, not a separate professional API plan; Pro-tier consumer output is treated the same as Free-tier output for watermark purposes. Third, this table is only about the visible mark. It says nothing about whether SynthID is present, because SynthID is present on all four rows.

Does every tier get SynthID?

Yes. Google states that SynthID applies across these tiers, including Ultra and API output that never receives a visible mark. The tier difference is only whether a visible corner mark is layered on top of SynthID, not whether the invisible signal exists underneath it.

That is worth sitting with, because it is the part most comparison pages get backwards. A "no watermark" tier is not a "no provenance signal" tier. SynthID is described by Google DeepMind as a signal embedded into the pixels themselves during generation, engineered to keep surviving common edits such as cropping, filters, resizing, and lossy re-compression. It is not a logo you can point to, and no tier lets you opt out of it. For the deeper mechanics of how that differs from a visible logo, and why the two get evaluated in completely different ways, read visible Gemini watermark vs SynthID.

How much does Google AI Ultra cost?

This guide anchors on tier behavior instead, because that part is verifiable straight from Google's own product documentation and does not shift with region or promotion. If price is the deciding factor for you, that is itself useful information: a free local tool that processes an existing Free- or Pro-tier image may cost you nothing and take less time than comparing subscription tiers in the first place.

Does using Ultra or the API mean the image has no watermark at all?

No. It means no visible corner mark is added. SynthID's invisible signal is still present on Ultra and API output by Google's own description, and this project does not detect or remove SynthID under any tier. "No visible mark" and "no watermark" are not the same claim, and treating them as interchangeable gives you a false sense of what an Ultra- or API-sourced image actually carries.

This distinction matters for disclosure obligations too. Rules such as the EU AI Act's Article 50 and California's SB 942 are concerned with machine-readable AI content signals, not specifically with a visible logo. An image with no visible mark can still be squarely in scope for those rules because SynthID, or a comparable machine-readable signal, is still attached to it. See why does Gemini put a watermark on my image for how the transparency reasoning behind the visible mark connects to that broader disclosure picture.

What to actually do with this

If you generate through Free or Pro and need the visible mark gone from an existing image, that is the case the local browser eraser is built for: it validates a supported visible-mark candidate on your device before changing anything, and it fails closed rather than guessing. If you can choose the generation surface up front, switching to Ultra or the API avoids the visible mark entirely for new images, at whatever cost or access tradeoff applies to that surface. The how-to guide walks through both routes side by side, google-ai-studio-watermark covers the API and AI Studio case specifically, including why this tool has nothing to do there, and gemini-watermark-sizes documents supported output dimensions as a separate question from tier.

Frequently asked questions

Which Gemini tiers show a visible watermark?

Free and Google AI Pro outputs carry both a visible corner mark and SynthID. Google AI Ultra and the AI Studio and API surfaces carry SynthID only, with no visible mark.

Does every tier get SynthID?

Yes. Google states SynthID applies across these tiers. The difference between tiers is whether a visible corner mark is layered on top of it, not whether the invisible signal exists.

How much does Google AI Ultra cost?

Unverified as of this draft. Public sources disagree, with figures including $249.99, $100, $99.99, and $19.99 per month depending on region, promotion, and source. Confirm the current price on Google's own pricing page and timestamp whatever figure is used before publishing it.

Does using Ultra or the API mean the image has no watermark at all?

No. It means no visible corner mark is added. SynthID's invisible signal is still present on Ultra and API outputs by Google's own description, and this project does not detect or remove SynthID.