OpenAI has quietly updated ChatGPT to reject requests that directly imitate the distinctive writing styles of well-known authors. This seemingly small change reflects the company's broader effort to navigate mounting copyright challenges.
Users who ask ChatGPT to write "in the style of" authors like Stephen King, J.K. Rowling, Ernest Hemingway, or Charles Dickens no longer receive text designed to closely replicate those authors' voices.
Instead, the chatbot explains that it can create an original piece inspired by the broader literary qualities associated with that writer while maintaining its own voice.
The update, first reported by Ars Technica, wasn't accompanied by a formal product announcement. Even so, it represents an example of how legal pressure is beginning to shape the behavior of commercial AI models.
A small change with larger legal implications
OpenAI’s timing is notable. The company continues to defend itself against multiple copyright lawsuits filed by authors who argue that its AI models were trained on copyrighted books without permission.
Several of those cases also point to ChatGPT's ability to generate text that closely resembles protected works. With this move, OpenAI may be attempting to reinforce the distinction between inspiration and infringement.
The new behavior also draws a clearer line between creative inspiration and direct stylistic imitation.
By steering users away from recognizable voices and toward broader literary characteristics, OpenAI may be trying to reduce one area of potential legal exposure while preserving the chatbot's applications as a creative writing assistant.
The change also aligns with guidance from industry groups such as the Authors Guild, which has encouraged writers to avoid using generative AI to deliberately imitate another author's distinctive voice for commercial purposes.
Newly limited creative writing capabilities remain
Users who request the distinctive style of a well-known author may not receive an outright refusal. Instead, ChatGPT may offer an original approach based on broader creative traits.
For instance, instead of writing exactly like Stephen King, the model may offer an original horror story built around atmospheric tension, character development, and small-town unease without attempting to reproduce King’s signature prose.
The method mirrors OpenAI's existing safeguards for image generation, with DALL·E having long declined requests to generate images in the style of living artists. Although OpenAI had not previously implemented similarly visible restrictions for text generation, together these moves show a broader effort to address intellectual property concerns across multiple AI products.
For writers, marketers, and enterprise users, the update shows how generative AI platforms are becoming increasingly governed around acceptable creative use cases rather than serving as unrestricted content generators.
Furthermore, those using generative AI for content creation should expect model behavior to continue evolving as vendors respond to legal and regulatory developments.
AI companies take different approaches
OpenAI isn't the only AI platform facing this issue, but the company’s competitors haven't adopted the same approach.
Research cited by Ars Technica found that Perplexity similarly refuses requests to closely imitate authors' writing styles, while Anthropic's Claude and Microsoft's Copilot have historically been more willing to fulfill those requests, often with disclaimers.
Google's Gemini has reportedly been more permissive with style-based requests, although model behavior may continue evolving as vendors update their safety policies.
That inconsistency presents a challenge for organizations building products on top of foundation models. A writing workflow or customer-facing feature that functions one way today could switch up tomorrow as providers update their policies or respond to court rulings.
For enterprises adopting generative AI, the latest change serves as another reminder that foundation models are not static platforms. As copyright lawsuits work their way through the courts and regulators continue examining generative AI, vendors are increasingly redefining what their systems will and won’t do.
For organizations relying on these platforms, staying current on those evolving guardrails is becoming just as important as keeping up with new model releases.
Also read: Grammarly faces a class action lawsuit alleging its Expert Review feature used real writers’ names to present AI-generated feedback without their consent.