OpenAI Trials ‘Writing Style’ Tool That Learns From Users’ Apps

OpenAI has begun testing a new capability for ChatGPT that allows the model to adapt its output to match a user’s personal writing style by analyzing samples drawn from the user’s linked applications.

The feature, dubbed “Writing Style,” works by scanning documents, emails or other text the user has stored in connected services, then using that data to fine‑tune the model’s phrasing, tone and structure for future interactions. The approach is intended to make the assistant’s suggestions feel more natural and aligned with the way an individual typically communicates.

OpenAI’s move follows a broader trend among AI developers to offer more customizable experiences. Earlier iterations of ChatGPT introduced “Custom Instructions,” letting users set preferences for tone or detail level, but those settings relied on explicit user input rather than automated learning from existing content. By leveraging real examples from a user’s own files, the new tool promises a deeper level of personalization without requiring manual configuration.

Privacy considerations are central to the rollout. OpenAI has stated that the analysis occurs locally on the user’s device where possible, and that any data used to train the style model is not retained beyond the session unless the user opts in to persistent learning. Nonetheless, the prospect of an AI system reading personal documents has sparked discussion among security experts, who caution that even temporary access could expose sensitive information if not properly sandboxed.

Potential applications range from drafting emails that sound like the sender to generating reports, social media posts or code comments that reflect a developer’s typical phrasing. For professionals who spend considerable time on written communication, the feature could reduce repetitive editing and speed up content creation, while still preserving a consistent voice across different platforms.

OpenAI has not disclosed a definitive launch date, but the testing phase appears to be limited to a subset of ChatGPT Plus subscribers. As the company gathers feedback, it will likely refine the balance between personalization and privacy, and may expand the feature to broader user groups if the response proves favorable.