Supposedly, humans prefer generated content.

It would be super cool if "journalists" writing about studies would be interested in at least attempting to assess the validity of a study by letting me know important details like... sample size, methodology, and confidence intervals.

For example, this study employed 10 professional writers vs. Chat GPT-4. Sample size for their survey participants was pretty good at 1,203 individuals responding to a randomized online survey. P-values are mostly insignificant, but there are a few standouts.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4453958

Anyway, it's interesting, but science reporting mostly sucks.