How AI Is Changing Language: Debates Over LLMs in Literature and Journalism

The literary and media worlds have been rocked by allegations of the covert use of large language models in the writing of texts. This applies not only to journalistic pieces but also to fiction—authors are suspected of having ChatGPT or other similar systems do some of the work for them.

Linguists interviewed by The Guardian are trying to pinpoint exactly what distinguishes a text written by a human from one generated by an LLM. According to their observations, machine-generated writing is often betrayed by overly polished phrasing, predictable sentence structure, and the absence of those rough edges and unexpected turns of phrase that are characteristic of a living author’s voice. At the same time, there are still no clear-cut, reliable markers that allow us to confidently distinguish machine-generated text from human-written text—the models are becoming increasingly convincing.

What Writers Are Saying

Well-known fiction writers have joined the discussion, including Jennifer Egan and Janet Winterson. They reflect on how the spread of generative AI will affect the future of fiction: whether the very concept of authorship will change, whether the value of human imperfection in a text will remain, and how readers will perceive books if they cannot be sure who—or what—wrote them.

Why This Matters

The debate goes beyond the technical question of “can a machine write like a human?” It touches on broader themes:

• readers’ trust in the printed and digital word;
• the status of writing and copyright in the age of generative models;
• changes in language itself—from journalistic clichés to literary devices—under the influence of texts increasingly created or edited by AI.

While the industry searches for ways to distinguish machine-generated text from human-written text, the language models themselves continue to learn from an ever-growing corpus of human-written texts—and, according to the publication’s interviewees, this line will become increasingly blurred.

Source: Hacker News - Newest: “AI” “LLM”