- Published on
How To Write With An LLM
- Authors
- Name
- DeepHorus
Thomas Ptacek, a prominent security researcher and blogger, published a new post on 17th September 2026 titled "How To Write With An LLM," arguing that writers should treat large language models as copyeditors rather than writing assistants. His central rule is blunt: writers may not use a single word an LLM suggests. Ptacek describes this as a form of intellectual personal protective equipment, urging strict adherence to keep AI-generated phrasing out of finished work.
Ptacek explains that he will not let LLMs write content for his blog, but he does use them for fact-checking, spelling and grammar, and occasionally as a thesaurus. He includes a proofreading prompt to help readers build their own copyediting tool. The post also features a screenshot of his personal LLM copyediting setup, along with a link to a related Twitter thread.
The reasoning behind the rule is both stylistic and practical. Ptacek notes that text containing LLM-suggested phrases has a distinctive "weird smell" that undermines authenticity. By banning those phrases outright, he says, writers can preserve their own voice and avoid the homogenized tone common in AI-assisted prose.
The implications extend beyond personal blogging. As LLM tools become embedded in everyday writing workflows, Ptacek's approach offers a disciplined alternative to full AI authorship. It positions the technology as a utility for verification and polish, not as a creative partner.
Ptacek's post links to recent articles on his blog, including "Generating running routes with GPT-6 Astra and ChatGPT Work" and "OpenAI agents attacked RubyGems back in May," both dated 12th September 2026. These pieces show his continued engagement with AI-related topics while maintaining his own editorial standards.
Looking ahead, Ptacek's copyeditor model may resonate with writers seeking to use AI without losing their identity. His prompt provides a practical starting point for others to adopt similar safeguards. The broader debate over how much AI should influence human writing is likely to intensify, but Ptacek has staked out a clear and disciplined position.
