OpenAI is testing a new ChatGPT feature designed to make AI-generated writing sound more like the user. Called Writing Style, the feature appears to use examples from a user’s existing messages, documents and emails to build a more personalised writing style.
The test could reduce the need for users to explain how they want ChatGPT to write repeatedly. Rather than relying on instructions such as asking for a casual, professional or conversational tone, the chatbot could study a user’s own writing and use those patterns when generating new content.
How ChatGPT’s Writing Style feature works
The feature was spotted during an onboarding process by X user Gael Breton. According to the screen shared online, ChatGPT can draw writing examples from three categories: Messaging, Documents and Email. Users can connect services such as Slack for messages, Google Drive or Notion for documents, and Gmail for email.
Is this new in ChatGPT? pic.twitter.com/DJfIWKjy4a
— Gael Breton (@GaelBreton) September 4, 2026
After connecting those services, users can apparently select the option to use their writing style. ChatGPT can then reference the user’s existing writing when producing new content. This could allow the system to identify characteristics such as sentence structure, vocabulary, level of formality and overall tone.
The approach could be particularly useful for people who regularly use ChatGPT to write emails, reports, messages or other professional material. Instead of providing several examples or a detailed style prompt each time, users could allow their existing writing to provide the necessary context.
For example, someone who normally writes short, informal workplace messages could have ChatGPT reproduce that style when drafting a new message. A user who writes longer, more formal emails could likewise give the system a larger sample to learn from.
The feature is still being tested, however, and the information currently available does not indicate exactly how much data ChatGPT uses or how it processes the writing samples. It is also unclear whether users can select individual sources or control which types of writing are used.
How it compares with Claude’s Styles feature
OpenAI’s approach is similar to a feature Anthropic previously offered for Claude. Claude’s Styles allowed users to provide writing samples by uploading or pasting text, after which the system could create a custom style based on those examples.
Anthropic later deprecated the standalone Styles feature and incorporated similar functionality into Claude Skills. That change made the process less straightforward for users who wanted an AI assistant to imitate their preferred writing style.
ChatGPT’s reported approach could therefore offer a simpler alternative. Rather than requiring users to prepare samples manually, the feature appears to be designed to draw from writing that already exists in connected services.
That could also give ChatGPT access to a broader range of examples. A collection of Slack messages could show how a person communicates informally, while documents could demonstrate how they structure longer pieces of writing. Emails could provide additional examples of how the same person adjusts their tone for more formal communication.
Using several types of writing may help an AI system develop a more complete picture of a person’s habits. It could distinguish between how someone writes a quick message to a colleague and how they prepare a formal document.
At the same time, connecting multiple third-party services raises questions about privacy and user control. Writing can contain personal, professional, and confidential information, so any system that analyses those sources will need to make clear what information it accesses and how it uses it. OpenAI has not yet provided full public details about those aspects of the reported test.
When could the feature arrive?
OpenAI has confirmed that it is testing the Writing Style feature, but there is currently no confirmed date for a broader release. As with other experimental ChatGPT features, the company could change how the system works before making it available to a wider group of users.
If released more widely, Writing Style could address a common complaint about AI-generated content: it can sound polished but still feel generic. Even when users provide detailed instructions, generated text can sometimes have recognisable patterns that make it feel different from their natural writing.
A system that learns directly from a person’s previous work could make that distinction less noticeable. It could also make ChatGPT more useful for recurring tasks, especially for users who frequently draft communications or documents with AI help.
The development reflects a wider shift in generative AI towards more personalised assistants. Instead of providing the same broadly defined experience to everyone, AI companies are increasingly looking at ways to adapt their tools to individual users, their workflows and their preferences.
For ChatGPT users, the potential benefit is straightforward: less time spent explaining how something should sound and more time spent reviewing the actual content. However, the feature’s success will depend on how accurately it captures a person’s voice while giving users enough control over the information used to build that profile.
For now, Writing Style remains a test, not a feature available to everyone. If OpenAI proceeds with a wider launch, it could give ChatGPT a more practical way to produce writing that feels personal rather than like standard AI-generated text.
