A Strict Rule for Writing with LLMs Without Losing Authentic Voice

Thomas Ptacek and Simon Willison advocate for a strict rule when writing with LLMs: use the models solely for copyediting and proofreading, but never use the exact phrasing suggested by the model. As AI-generated text becomes ubiquitous, maintaining an authentic human voice and intellectual discipline is increasingly critical for writers. This approach allows creators to benefit from AI feedback and proofreading without letting generic AI phrasing dilute their personal tone. Under this framework, LLMs are limited to tasks like spellchecking, grammar review, fact-checking, and serving as a thesaurus. Ptacek also shared a custom prompt and screenshot of his personal LLM copyediting tool to help others construct a similar workflow.