OpenAI Tests Writing Style Feature That Learns From Connected Apps

OpenAI appears to be testing a feature called "Writing Style" that could allow ChatGPT to learn and mimic a user's natural writing voice by referencing content from their connected apps.

An onboarding screen for the feature states: "ChatGPT will write in your voice by referencing examples from your connected apps." Users can provide source material across three categories: messaging, documents, and email. The supported services listed include Slack for messaging, Google Drive and Notion for documents, and Gmail for email. Once configured, ChatGPT would reference the user's natural writing patterns from those services when drafting new content, rather than requiring users to repeatedly explain their preferred tone or paste examples into each conversation.

The feature bears similarity to Anthropic's earlier "Styles" personalisation tool, which allowed users to upload or paste writing samples to generate a custom style. However, OpenAI's approach goes further by letting ChatGPT reference writing that already exists inside connected apps, potentially giving the model a larger and more diverse set of real examples — including how a user writes emails compared with messages or longer documents.

OpenAI has confirmed it is testing Writing Style but has not announced a timeline for wider rollout. The feature arrives amid growing competition among AI providers to offer more personalised, context-aware writing assistance.

Why This Matters

If rolled out widely, the feature could save users significant time by eliminating the need to repeatedly specify tone or style preferences when asking ChatGPT to draft content.
Connecting ChatGPT to personal apps like Gmail and Slack raises privacy and data security questions, as the model would need access to potentially sensitive communications to learn writing patterns.
The development signals that personalisation is becoming a key battleground among AI assistants, with OpenAI moving beyond simple style selection toward deep integration with users' existing digital workflows.

Background

AI writing assistants have long offered basic tone controls — formal, casual, persuasive — but true personalisation requires learning an individual's unique voice. Anthropic's Claude was among the first to offer custom "Styles" that users could train by pasting or uploading samples. OpenAI's experiment takes that concept further by pulling examples directly from apps the user already trusts with their data. The move aligns with a broader industry push toward AI agents that can operate across multiple services on a user's behalf.

Key Perspectives

Users seeking productivity gains: Many welcome the idea of an AI that "gets" their voice without manual setup, particularly for routine writing tasks like emails and messages. Privacy and security advocates: They are likely to raise concerns about granting ChatGPT access to personal Slack messages and emails. Questions about how sample data is stored, used for training, and whether it can be deleted will need clear answers. Competitors: Anthropic and Google (with Gemini) are working on similar personalisation features. OpenAI's direct integration with connected apps could set a new user expectation, forcing rivals to match or differentiate.

What to Watch

OpenAI's privacy policy updates and any opt-out mechanisms for users who do not want their app data used for style learning.
Rollout timeline: whether this remains a limited test or expands broadly in the coming weeks.
Competitor responses: Anthropic may deepen Claude's app integrations; Google could leverage its Workspace ecosystem for similar functionality.

ZOTPAPER