Steelman · slot C
The collapse-of-textual-trust case
An essayist navigating a post-LLM readership would argue —I now hesitate before an em dash. That's the small, stupid, telling fact of writing in 2026. Mahadevan admits he second-guesses his own sentences wondering if readers will assume AI wrote them; Bender says people send her synthetic text she can't reliably identify. When 32% of small businesses route customer communication through LLMs and a polished paragraph carries no presumption of a human author, every writer becomes a suspect and every reader becomes a forensic analyst. The damage isn't just to AI-generated text — it's to genuine voices, who must now either rough up their prose to prove they exist or accept being mistaken for a chatbot. Trust in authorship is a public good, and it's quietly being spent down.