[MODEL] Claude’s peculiar way of writing is spreading.

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Other unexpected behavior

What You Asked Claude to Do

Claude’s peculiar way of writing is spreading. I suspect that a form of “watermarking” is behind it. This has not been officially established; it is my hypothesis.

The problem is that these verbal habits do not stay within Claude. Its output is turned into documents, which become the basis for further writing, spreading the same unusual vocabulary and expressions. People read these texts to learn about a subject and begin picking up the peculiar prose along with the information.

Should that seem abstract, do take a look at the feeds on LinkedIn and similar platforms. To my eye, even human-written posts are already acquiring Claude’s verbal habits. Whether a human or an AI wrote them is no longer the central question. The point is that both are spreading the same language.

I see similar vocabulary even in other models suspected of unauthorised distillation from Claude. One might have hoped those particular habits did not require quite such faithful reproduction.

This is not simply a matter of disliking a writing style. If the spread continues, people may increasingly speak and write like Claude—not just in online feeds, but in everyday life. One model’s peculiar verbal habits could displace other forms of human expression and leave a permanent mark on language itself. That linguistic contamination is what I consider a serious threat.

What Claude Actually Did

Claude repeatedly uses unnatural vocabulary and formulaic phrasing. In my observation, similar patterns are spreading to other models’ outputs and human-written text.

Expected Behavior

Claude should produce natural, context-appropriate language. Please investigate and address the cause of these patterns. If watermarking is involved, it should not operate in a way that distorts human expression.

Files Affected

Permission Mode

Accept Edits was ON (auto-accepting changes)

Can You Reproduce This?

Yes, every time with the same prompt

Steps to Reproduce

_No response_

Claude Model

Opus

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

2.1.277

Platform

Anthropic API

Additional Context

_No response_