ChatGPT blocks direct requests to imitate named authors

OpenAI’s ChatGPT is now declining requests to write in the exact style of well known authors, instead offering text based on broader characteristics such as mood, genre and narrative approach. Kyle Orland reports for Ars Technica that the change affects prompts referring to both living and deceased writers, including Stephen King, J.K. Rowling, Amy Tan, Charles Dickens and Ernest Hemingway.

When asked to produce a horror opening in King’s style, ChatGPT said it could write atmospheric, character driven horror with small town dread, but could not closely imitate the author’s distinctive voice. It then generated an original alternative intended to create a similar effect.

The distinction may matter in ongoing legal disputes over AI training and output. US copyright law generally protects the concrete expression of a work rather than an author’s general style. However, an imitation could still create legal risks if generated text is substantially similar to protected material.

Authors have raised ethical and commercial concerns as well. The Authors Guild advises writers not to use generative AI to deliberately copy another writer’s voice or distinctive traits, particularly where such use could harm the value of the original work or generate profit.

Other AI providers take different approaches. According to a study by No Latency cited by Ars Technica, Google’s Gemini complied with author style prompts, while Perplexity redirected users. Anthropic’s Claude and Microsoft Copilot reportedly complied in some cases while adding cautions about imitation.