Re:The Science of Detecting LLM-Generated Text
by slopinthebag on Sunday March 01, 2026 @04:28AM
(#47205132)
I'm not so sure I buy that. AI written text is fairly obvious to good writers with exposure to LLM output. Is it a case where it's sort of an average of writing styles, but that average is not human and thus humans can detect it?
[ Parent
| on HN ]
AI writing you can recognize as AI writing is obvious. Newer models are better about this and the line will only get more blurry. Here's a benchmark where good writers make the assessment rather than different LLMs ranking each other: https://surgehq.ai/leaderboards/hemingway-bench
The top models are also the latest:
Gemini 3.1 Pro: still a bit of a gremlin, but will probably stay on top until the other model makers go xkcd 810 and target this benchmark
Gemini 3 Flash: current favorite of writers using it as a helper for its speed and decent prompt following
Re:The Science of Detecting LLM-Generated Text
by slopinthebag on Sunday March 01, 2026 @04:30PM
(#47210866)
Yeah I think it's more about effort than anything - if the user puts in effort to make the writing indistinguishable from human writing, I'm not so sure it's really a bad thing. Low effort slop is detectable however, and that's a good sign to just not continue reading it.