Kenneth Glad
A hotel review, a short story, a news sentence with a repeated word. In the age of chatbots, ordinary writing can suddenly look suspicious. The uncertainty is changing how people judge language, authorship and trust.
Claire Hardaker, a forensic linguistics professor at Lancaster University, runs Bot or Not, an online test that asks people to judge whether short pieces of writing were produced by a person or by AI.
In an example cited by The Guardian, the test used hotel reviews, asking readers to separate authentic comments from machine-generated ones. Users identified the AI-written reviews correctly only about 60% of the time, showing how difficult that judgment has become.
Many rely on quick warning signs, including cliches, dashes or neat groups of three. But those patterns also appear in ordinary human writing.
Hardaker says the clues are unreliable: “People have learned very simplistic rubrics and now just madly apply them everywhere.”
Wrongly accused
The danger is not limited to machine-generated writing slipping through unnoticed. A real writer can also be misidentified, and an online accusation can spread faster than evidence or context.
That can leave authors, journalists and publishers defending ordinary choices of style, punctuation or wording. A sentence that once looked merely polished may now be treated as suspicious.
Large language models are trained on vast collections of human writing. At the same time, readers and writers are seeing more AI-generated style in emails, searches and chatbot replies.
The result is a feedback loop in which human and machine writing increasingly resemble each other. That makes certainty harder, especially when judgments are based on instinct.
Media outlets, including The Guardian, now receive complaints about supposedly machine-generated writing.
Delve and doubt
Researchers cited by The Guardian have linked chatbot use to increased use of words such as “delve”, “showcase” and “intricate.”
Those terms have appeared more frequently in research papers and other writing since generative AI became widely available.
Other studies suggest AI editing can pull English toward a more standard Anglo-American form, reducing some regional or cultural variation.
Researchers have described this trend as “cultural ghosting,” where local expressions and established writing conventions are gradually replaced by more uniform alternatives.
The shift does not necessarily make writing less accurate or easier to understand. However, it may make different voices sound more alike over time.
So the question is no longer just whether a sentence sounds polished. It is whether voice, background and intent are being quietly narrowed as AI-generated language becomes part of everyday communication.
Fiction’s bargain
Gary Shteyngart, a Soviet-born American novelist and professor of writing at Columbia University, told The Guardian that his students reacted angrily when one classmate planned to use AI in fiction.
Shteyngart, known for satirical novels including Super Sad True Love Story and the memoir Little Failure, said the reaction reflected a deeper concern about trust between writer and reader.
“Reading literary fiction is this incredible Vulcan mind meld with another human being,” he said.
Jennifer Egan, the Pulitzer Prize-winning American novelist best known for A Visit from the Goon Squad, has rejected the use of generative AI in her own creative work, arguing that writers should develop their craft rather than rely on the technology.
By contrast, British novelist Jeanette Winterson, author of Oranges Are Not the Only Fruit and a longtime commentator on the relationship between technology and art, takes a more open approach. She sees AI as another creative tool, but argues that it cannot replace human feeling, lived experience or emotional understanding.
Peter Stockwell, professor of literary linguistics at the University of Nottingham, makes a related argument from another angle. He told The Guardian that AI can produce clear and functional text, but not great storytelling.
“If you want something that’s very familiar and very mediocre and entirely functional, it’s amazingly good at that.”
Source: The Guardian, Bot or Not
