How AI is changing language
TL;DR
The Guardian uses hotel reviews to show how weak human AI detection still is. Linguist Claire Hardaker says her Bot or Not test lands at roughly 60% accuracy; common tells such as cliches, dashes and neat groups of three also appear in ordinary human writing. The piece links recent literary and media disputes, including allegations around Jamir Nazir, the withdrawn novel Shy Girl and Steven Rosenbaum's book with hallucinated quotations. The bigger story is a culture of suspicion, not just whether a text used AI.
Nauti's Take
The sharpest point in the Guardian piece: AI does not merely make language flatter, it makes readers more suspicious of language. That is more dangerous than a few overused words.
Strong editorial teams should stare less at detector traffic lights and look harder at provenance, intent, texture, sourcing and lived experience. The new quality edge is recognizable judgment, not model prose scrubbed until it looks human.
Briefingshow
The real change is not one suspicious sentence, but the feedback loop between models, users, editors and detectors. Once people start writing against supposed AI markers, human language shifts too. For content teams, detector scores are not proof; style work becomes a strategic editorial discipline.
Sources