---
title: "How Claude's text watermarking works"
description: "Anthropic describes a text-watermarking system planned for Claude to estimate whether Claude was involved in generating or processing text, primarily to comply with the EU AI Act. The method uses a key and preceding words to influence the source of randomness in low-stakes word choices, creating a detectable pattern without adding characters, tokens, identifying information, or extra cost. Anthropic says internal testing found no impact on content, creativity, readability, or perceived quality, while detection confidence improves with longer passages and weakens when text offers few valid alternatives, such as factual statements, proofreading, or code. The watermark can indicate only that Claude was likely involved, not whether text was human-written, who authored it, or which person or organization used Claude; light editing may not remove it, but a complete rewrite will."
---

# How Claude's text watermarking works

[Anthropic](https://yomu.fyi/company/anthropic) · Sep 1, 2026

**Type:** Explainer

## Summary

Anthropic describes a text-watermarking system planned for Claude to estimate whether Claude was involved in generating or processing text, primarily to comply with the EU AI Act. The method uses a key and preceding words to influence the source of randomness in low-stakes word choices, creating a detectable pattern without adding characters, tokens, identifying information, or extra cost. Anthropic says internal testing found no impact on content, creativity, readability, or perceived quality, while detection confidence improves with longer passages and weakens when text offers few valid alternatives, such as factual statements, proofreading, or code. The watermark can indicate only that Claude was likely involved, not whether text was human-written, who authored it, or which person or organization used Claude; light editing may not remove it, but a complete rewrite will.

## Context

Anthropic says it is implementing watermarking to comply with the EU AI Act, which requires AI system providers to mark AI-generated text. The change is being applied globally at launch because Anthropic does not yet have a durable way to scope it by region.

## Approach / What changed

The system uses a version of Google DeepMind’s SynthID-Text approach. A key and several preceding words determine the randomness used for eligible word choices, allowing a detector with the key to estimate whether Claude generated the passage without adding visible or hidden content.

## Takeaways

- Watermarking changes the source of randomness for low-stakes word choices rather than forcing Claude toward particular words or adding hidden characters.
- Detection is less effective for short passages and text with few flexible choices, including factual statements, lightly edited human writing, proofreading, and code.
- The detection API can indicate likely Claude involvement, but cannot establish human authorship, identify a person or organization, or distinguish writing from heavy editing.

**Tags:** [AI](https://yomu.fyi/topic/ai), [Compliance](https://yomu.fyi/topic/compliance), [Generative AI](https://yomu.fyi/topic/generative-ai), [Watermarking](https://yomu.fyi/topic/watermarking)

- Source: [Anthropic](https://www.anthropic.com/news/claude-text-watermark)
- Source URL: https://www.anthropic.com/news/claude-text-watermark
- Ingested by Yomu: 2026-09-09T11:35:12.821Z

[Read original post](https://www.anthropic.com/news/claude-text-watermark)