ChatGPT mimics your writing via app data

OpenAI appears to be testing a new “Writing Style” feature for ChatGPT that could let the AI assistant learn how you write by pulling examples from your connected apps. The feature, currently available to a small group of users, is designed to generate text in your voice without requiring you to spell out your preferred tone each time.

Feature learns from your apps

An onboarding screen shown to testers says, “ChatGPT will write in your voice by referencing examples from your connected apps.” Users can provide examples across three categories: Messaging, Documents, and Email. Under these categories, OpenAI lists Slack, Google Drive and Notion, and Gmail as example sources.

Once configured, ChatGPT could reference the way you naturally write across those services when drafting new content. That means instead of pasting writing samples into a conversation or explaining your style preferences every time, the model could pull from your existing communications to mimic your patterns.

Deeper with connected services

This approach goes further than a similar personalization feature Anthropic previously offered called Styles. Anthropic’s Styles let users upload or paste writing samples, and ChatGPT’s experiment appears to let the model draw on writing that already exists inside connected services.

That could give ChatGPT a much larger set of real examples, including how you write emails compared with messages or longer documents. By tapping into a broader range of your writing, the model might develop a more accurate sense of your style across different contexts.

Rollout and availability unclear

OpenAI has confirmed it’s testing Writing Style, but the company has not shared details on when it will reach all users. The test is limited for now, and it remains unclear which users have access or how widely the feature will expand.

While no timeline has been announced, the feature appears to be in an early experimental stage. OpenAI has not commented on potential privacy implications or how user data from connected apps would be handled, and how the company stores and uses this data will likely be a point of scrutiny for privacy advocates.

Similarities to Claude Styles

The feature bears a resemblance to Claude Styles, which allowed users to generate a custom style based on their communication patterns. OpenAI’s version, however, leverages connected apps, potentially giving ChatGPT access to a much larger corpus of your writing.

That difference could be significant. If you write casually in Slack but more formally in Gmail, the model could learn to adapt accordingly, perhaps offering more nuanced tone matching than a static style profile.

What this could mean for users

This testing suggests OpenAI is pushing toward deeper personalization, making AI assistants more attuned to individual writing habits. While convenient, it also means AI tools may hold more personal data than before.

The balance between custom functionality and user privacy is delicate, and the outcome of this experiment could set a precedent for how AI services handle personal content. As the feature evolves, users will likely want to know exactly what data is accessed, how long it’s retained, and whether they can opt out.

Why it matters

The push to personalize AI through app data raises real questions about data access and trust. If ChatGPT can scan your emails and messages, it needs to handle that content with strong safeguards, but so far OpenAI has not detailed those protections. For users, this feature could offer convenience, but it also means sensitive communications might feed into the AI’s memory. As tests progress, the key will be whether OpenAI can balance personalization with clear, transparent data practices that users can trust.