Claude watermark detector

AI Watermark Center cannot currently verify Claude's statistical watermark because Anthropic's detection method requires provider-side detection capability that has not yet been publicly released for general use.

Current architecture status for a future official integration: provider detection capability is Announced, not generally released; AI Watermark Center integration is Not integrated. Announced is not the same as a public detector, and neither is integration. Those fields exist so a public Anthropic detection API can be wired in later without pretending it exists today.

What our text scanner can inspect

• Hidden Unicode
• Zero-width characters
• Formatting anomalies and normalization differences

What it cannot inspect

The AI Text Watermark Scanner cannot inspect Claude’s statistical watermark. It inspects hidden Unicode, zero-width characters, and formatting anomalies only.

Anthropic says detection depends on a key, and that it will soon offer a watermark detection API. Implementation details were still forthcoming when these sources were checked. Until that capability is public and independently integrable, this page will not grow a fake score, a paste box labelled “Detect Claude Watermark,” or a third-party detector ranking.

Anthropic states that Claude’s official text watermark is not based on hidden characters. This scanner checks Unicode artifacts only.