How AI is changing language
TL;DR
The Guardian uses hotel reviews to show how unreliable AI spotting has become: forensic linguist Claire Hardaker says many people score only about 60 percent in detection tests. Common tells such as cliches, em dashes or neat groups of three are weak evidence, because they also appear throughout human writing. Research points to LLM patterns: overused words like delve, showcase and underscore, more nouns, fewer pronouns and a pull toward Anglo-American standardisation.
Nauti's Take
The real issue is not that AI produces bad writing. It is that it normalises smooth, average language at scale, making human writing more cautious, more uniform and more suspect.
Writers should spend less energy playing AI detective and more energy building friction: concrete observation, voice, risk and stance. That is still where the distance from the machine shows.
Briefingshow
The debate is moving from whether a text is AI-made to how AI is changing the way humans write. If detectors remain uncertain and people absorb machine-like patterns, authenticity becomes a question of social trust, not just technical classification.
Sources