How to spot AI writing

Large language models (LLMs) generate more than a third of new websites, and are used for essays and papers.

LLMs have stylistic quirks, such as using long em-dashes and words like 'maximise', 'deep dive', and 'delve'.

AI writing is not identified by single words, as that can be misleading; detection requires broader evidence.

Detection algorithms, like Pangram's, claim 99.98% accuracy but can give false positives without explanations.

Researchers compare texts from before and after LLMs, but it's hard to separate AI quirks from other language trends.