OpenAI has quietly changed how ChatGPT responds to requests that ask it to write in the exact style of well-known authors.
Instead of closely imitating a writer’s distinctive voice, the AI now declines such requests and offers original text inspired only by the author’s broader characteristics.
According to testing by Ars Technica, prompts asking ChatGPT to write like authors including Stephen King, J.K. Rowling, Amy Tan, Charles Dickens, and Ernest Hemingway now trigger a similar response. Rather than reproducing an author’s signature style, ChatGPT explains that it can capture general qualities—such as atmosphere, pacing, or tone—while creating content in its own original voice.
For example, when asked to write a horror story introduction in Stephen King’s style, ChatGPT responded that it could create an atmospheric, character-driven opening with small-town horror elements but could not imitate King’s exact writing style. It then generated an original passage with a similar mood instead.
The change could have important legal implications for OpenAI, which is facing multiple copyright lawsuits from authors who argue that its AI models were trained on copyrighted books without permission. One lawsuit specifically points to ChatGPT’s ability to generate text that closely resembles copyrighted works. OpenAI has not publicly commented on the latest change.
Under U.S. copyright law, an author’s general writing style is typically not protected, while the specific expression of ideas is. However, AI-generated text that becomes substantially similar to an author’s original work could still raise copyright concerns.
Legal experts say AI has made it easier than ever to imitate individual creative voices. George Washington University Law School professor Robert Brauneis told Bloomberg Law that creators have never before faced a situation where their personal writing styles could be copied so easily and at such low cost.
The Authors Guild has also advised writers not to use generative AI to intentionally copy another author’s unique voice or style. The organization says doing so may create ethical concerns and could potentially lead to claims involving unfair competition or copyright infringement.
Concerns over AI style imitation have surfaced before. Last year, author Lena McDonald received criticism after an apparent AI-generated editing note accidentally appeared in a published book, stating that a passage had been rewritten to better match another author’s style. Researchers have also published several studies examining how different large language models handle requests to imitate famous writers.
OpenAI already applies similar restrictions to image generation. Its DALL·E 3 model is designed to reject requests that ask for images in the style of living artists. However, OpenAI’s public documentation has not clearly outlined equivalent rules for written text, making the recent behavior appear to be a policy shift rather than a formally announced change.
Not every AI assistant handles these requests the same way. Research from No Latency found that Google’s Gemini generally complied with style imitation prompts, while Perplexity AI consistently refused and redirected users in a manner similar to ChatGPT. Anthropic’s Claude and Microsoft’s Copilot fell somewhere in between, generating stylistically similar content while acknowledging potential concerns about imitation.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
The change has frustrated some longtime ChatGPT users who relied on detailed style-based prompts for creative writing. Several users on Reddit say they now receive refusals where earlier versions of ChatGPT would generate text closely resembling the work of specific authors, forcing them to rethink how they structure their prompts.
