AI Text Watermarking by Model: Who Actually Does It
The assumption that every major model quietly watermarks its output is wrong, and the gap between assumption and reality is large. As of August 2026 exactly one provider runs a text watermark in production at scale.
This page is deliberately factual and dated, because it is a moving target. Where a provider has made no public commitment, we say so rather than guessing.
Google — Gemini: deployed
Google DeepMind's SynthID-Text is live in Gemini and is the first production text watermark deployed at this scale. Instead of the simple green-list bonus, it uses an approach DeepMind calls tournament sampling: candidate tokens are run through a bracket of pairwise comparisons scored by a keyed pseudo-random function, and the winner is emitted. The effect is similar — the output is biased toward tokens the key prefers — with better quality preservation.
Google reported no significant difference in user thumbs-up or thumbs-down rates across nearly 20 million Gemini responses between watermarked and unwatermarked output, which is the strongest public evidence that watermarking need not degrade quality.
Two limitations matter, and DeepMind states both:
- It weakens under paraphrasing, translation and heavy editing.
- The open-sourced implementation lets developers watermark their own models with their own keys. It does not let anyone verify Google's production Gemini output — that requires Google's key.
OpenAI — ChatGPT: built, never shipped
OpenAI developed a text watermarking prototype and did not deploy it. Reporting on the decision pointed to concerns about disadvantaging non-native English users, the ease of defeating it, and the effect on users who rely on the product.
Separately, OpenAI withdrew its AI Text Classifier — a detector, not a watermark — in 2023, citing low accuracy.
Practical position: assume ChatGPT output carries no watermark, while noting that this is a product decision that could change.
Anthropic — Claude: not deployed
Anthropic has not deployed a text watermark in Claude. There is no public API for detecting a watermark in Claude output, because there is no watermark to detect.
We are explicit about this because it is the most common misconception we see, and because tools that claim to "remove Anthropic watermarks" are selling something that does not currently exist. Our own site is built so that if an official detection API is ever released, enabling it is a configuration change — and until that day the interface says plainly that its scores are its own estimate.
Meta — Llama: no text watermark
Meta has published watermarking work for images and audio, and has released open-weight Llama models. Open weights make output watermarking largely unenforceable: anyone running the model locally controls the sampling loop and can simply omit it.
Open-weight models generally: structurally unwatermarkable
This is the point most coverage misses. Generation-time watermarking requires control of the sampling step. If you can download the weights and run them yourself, you own that step — so any watermark is opt-in for whoever operates the model. This is not a gap that will be closed; it follows from what open weights are.
Status at a glance
| Provider | Text watermark | Public detector |
|---|---|---|
| Google (Gemini) | Deployed — SynthID-Text | No — needs Google's key |
| OpenAI (ChatGPT) | Built, not shipped | No |
| Anthropic (Claude) | Not deployed | No |
| Meta (Llama) | No | No |
| Open-weight models | Operator's choice | No |
What follows from this
Most AI text in circulation is not watermarked. If a document was written with ChatGPT or Claude, there is no watermark in it, and no watermark detector will find one.
Accusations therefore almost never come from watermarks. They come from perplexity classifiers, which are far less reliable and carry a documented bias against non-native English writers. If you are contesting a flag, that distinction is your strongest argument.
A "no watermark detected" result proves very little. Absence of a signal is expected for most models. It is not evidence of human authorship.
Adoption could widen. Regulation such as the EU AI Act pushes toward machine-readable marking of synthetic content. If more providers deploy, this page will need revising — it reflects August 2026.
Sources
- Google DeepMind, SynthID and Watermarking AI-generated text and video with SynthID.
- Liang et al., GPT Detectors Are Biased Against Non-Native English Writers (2023).