How AI is changing language

Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Claire Hardaker, professor of forensic linguistics at the University of Lancaster, found that people correctly identify AI-generated hotel reviews only about 60% of the time in her 15-review 'Bot or Not' test, despite widespread confidence in their own ability.
- Common 'tells' people rely on—em dashes, the 'rule of three,' and stock phrases—are also characteristic of human writing, Hardaker notes, because LLMs were trained on it: 'You could go back to Charles Dickens and say he had AI, because he used the em dash too.'
- Words like 'delve,' 'showcase,' 'boast,' 'underscore,' 'garner,' 'align,' 'surpass,' and 'intricate' have been identified as 'focal words' LLMs overuse, with 'delve' first flagged after a researcher spotted its sudden spike in scientific papers.
- LLMs prefer attributive adjectives ('the uncomfortable chair') over predicative ones ('the chair was uncomfortable'), use nouns more than pronouns, and tend to homogenize global English toward Anglo-American standards in a process researchers call 'cultural ghosting.'
- Hachette withdrew debut horror novel Shy Girl after online rumors its author used AI (which she denies), and Steven Rosenbaum's The Future of Truth was found to contain hallucinated quotations he later apologized for.
- Detector Pangram claims a false positive rate of about 1 in 10,000, but the Guardian reporter fooled it on the first attempt; Hardaker, who has served as an expert witness, calls herself 'extremely sceptical' of such tools' efficacy.
- Studies found 'delve' and 'boast' spiked in unscripted human conversations after ChatGPT's release, while another showed 'delve' frequency in academic abstracts actually dropped after being called out on social media—evidence the influence runs both ways.
Why it matters: The erosion of reliable detection has already triggered real-world consequences: a debut novelist lost her publisher, a serious non-fiction book was exposed for hallucinated citations, and media outlets like the Guardian field mounting reader complaints—meaning writers, publishers, and educators are now operating in a regime where suspicion has replaced evidence, and even commercial detectors can be trivially fooled or falsely flag neurodivergent writers.



