Use an LLM to edit your own words, not to write new ones

When collaborating with an LLM on prose, keep the model in an editing role rather than a generation role: feed it your own draft and ask it to streamline and improve the language, then manually rework its output so it does not carry the telltale flatness of machine writing. This preserves the writer's voice and intent while still benefiting from the model's ability to catch clutter and awkward phrasing. The discipline is in applying a small number of explicit constraints rather than letting the model freely rewrite.

Source: https://sockpuppet.org/blog/2026/09/17/how-to-write-with-an-llm/