AI-sounding text is now a trust problem

'LLM Voice' (50290177) argues that if you want to reach humans, your humanity has to ooze out of your prose. One commenter says the grating sameness is the brain flagging that there's nothing to gain from the material, and another admits that reading so much LLM output makes them adopt its tics. An Ask HN (48886741) asks for a flag for AI-generated articles. Replies note HN's guidelines already ban generated comments, that AI detectors are unreliable, that voting can be gamed, and that on Lobsters the stigma got so bad some users felt unwelcome. A thread about Cliff Stoll's 'rumors of my death' (48037336) shows an AI-written obituary, with a commenter asking how we know he isn't an AI simulation of Cliff. A README in the Astroterm thread (50270651) earns praise with 'vibe coders, please take notes.'

The pattern: the cost of producing text hit zero, so readers are pricing in suspicion. The sharp part is that people can't reliably prove or disprove authorship, which means the penalty falls on anyone whose writing is merely polished.

Some commenters say many voters can't tell, or don't care if the premise is interesting.