Someday, the only answer will be to run everything you write online through an LLM to disguise your stylometric fingerprint.

https://arxiv.org/pdf/2602.23079

Recent advances have shown that LLMs can implicitly encode authorial style, enabling zero-shot or few-shot authorship identification without explicit training. Now with tool-augmented LLM agents, these capabilities can even be further amplified.