Thomas Ptacek Advocates Strict Rules for Writing With LLMs

The rise of large language models has led many professionals to rely on AI for drafting content, but prominent tech figures are now advocating for a much more disciplined approach. Security researcher Thomas Ptacek recently outlined a strict methodology for using these tools, arguing that writers should treat AI as a copyeditor rather than a creative assistant. His core directive, dubbed 'Rule Number One', dictates that writers must not use a single word or turn of phrase suggested by an LLM in their final text.

Ptacek suggests adopting this rule as a form of intellectual personal protective equipment. By strictly forbidding the use of AI-generated phrasing, writers can prevent their work from adopting the distinct, often formulaic style that characterizes machine-generated text. Instead of letting the AI write, authors should use the technology to identify weaknesses in their own prose, which they can then rewrite themselves.

Prominent developer and blogger Simon Willison endorsed this disciplined approach on September 17, 2026, noting that he refuses to let LLMs write content for his own blog. Instead, Willison limits his use of the technology to fact-checking, spelling, grammar correction, and acting as an occasional thesaurus. He agreed that avoiding LLM-suggested phrasing is an excellent principle for maintaining writing discipline and avoiding the recognizable 'weird smell' of AI prose.

To help writers implement this workflow, Ptacek shared details of his personal LLM-based copyediting tool, including a prompt designed to help others build similar systems. For practitioners, this shift in perspective changes how AI tools are integrated into the creative process. Rather than outsourcing the act of writing to a model, professionals can leverage LLMs to sharpen their editing process while ensuring their unique human voice remains entirely intact.