ARTFEED — Contemporary Art Intelligence

AI's Impact on Language and Literature: From Detection to Creation

ai-technology · 2026-07-06

A Guardian feature examines how AI is reshaping human language and literature, featuring insights from linguists Claire Hardaker and Peter Stockwell, and novelists Jennifer Egan and Jeanette Winterson. Hardaker's Bot or Not test shows people identify AI-generated hotel reviews only 60% of the time, while common 'tells' like em dashes and the rule of three are also human traits. LLMs overuse words like 'delve', 'showcase', and 'boast', but this may stem from human feedback training. Studies show 'delve' spiked in conversations after ChatGPT's release but dropped after social media scrutiny. Stockwell argues AI excels at lower linguistic levels but fails at narrative structure and originality, as it is trained on existing language and cannot innovate. Egan avoids AI entirely, fearing 'infection', and advises young writers to 'stay the fuck away'. Winterson sees AI as a tool but notes machines lack a limbic system and cannot convey inner realities. The article also covers controversies: Jamir Nazir's prizewinning story was suspected of AI use (he denied it), and Hachette withdrew Shy Girl after AI rumors. AI detectors like Pangram have low false positives but can be fooled by bombastic human writing. The piece concludes that human innovation, driven by social and bodily experience, remains AI's limitation.

Key facts

  • Claire Hardaker's Bot or Not test shows people identify AI-generated hotel reviews only 60% of the time.
  • Common AI 'tells' like em dashes and the rule of three are also characteristic of human writing, including Charles Dickens and Julius Caesar.
  • LLMs overuse words such as 'delve', 'showcase', 'boast', 'underscore', 'garner', 'align', 'surpass', and 'intricate'.
  • The 'delve' phenomenon may be due to reinforcement learning with human feedback, not the models themselves.
  • Studies show 'delve' frequency spiked in unscripted conversations after ChatGPT's release but dropped after social media attention.
  • Peter Stockwell says AI is good at lower linguistic levels but fails at narrative structure and originality.
  • Jennifer Egan avoids AI entirely, advising young writers to 'stay the fuck away'.
  • Jeanette Winterson views AI as a tool but notes machines lack a limbic system and cannot convey inner realities.
  • Jamir Nazir's prizewinning short story faced AI accusations; he denied using AI.
  • Hachette withdrew Shy Girl after online rumors of AI use, which the author denies.
  • AI detector Pangram boasts false positive rates of around 1 in 10,000 but can be fooled by bombastic human writing.
  • AI tends to flatten and homogenize language towards an Anglo-American standard, a process called 'cultural ghosting'.

Entities

Artists

Institutions

  • University of Lancaster
  • Columbia University
  • University of Nottingham
  • Hachette
  • The Guardian
  • The Atlantic
  • Anthropic
  • Pangram
  • Google

Locations

  • London
  • New York
  • Nigeria
  • Bay Area

Sources