OpenAI Tests ChatGPT Feature to Mimic User Writing Style via Connected Apps

OpenAI is testing a new "Writing Style" feature for ChatGPT designed to learn and replicate a user's unique communication patterns. The tool marks a significant shift toward automated, data-driven personalization by analyzing existing user data rather than relying on manual input.

According to BleepingComputer, the feature enables ChatGPT to build a stylistic profile by scanning writing examples directly from a user's connected personal applications. Confirmed integrations include messaging via Slack, documents in Google Drive and Notion, and emails through Gmail. By leveraging these connectors, the AI can identify the specific cadence, tone, and vocabulary a user employs across different platforms to generate more authentic-sounding output.

The Shift to Automated Personalization

Until now, personalization in large language models (LLMs) has primarily relied on "Custom Instructions" or manual prompting, where users provide a few specific examples of their work to guide the AI. Competitors have attempted to streamline this process; for instance, Anthropic introduced a "Styles" feature for Claude that allows users to tailor responses to a preferred voice, typically through manual samples or descriptive instructions.

OpenAI's approach evolves this concept by integrating style mimicry with its existing ecosystem of connected apps. Instead of requiring the user to curate and upload text samples, the system automates the gathering of training data from the user's actual digital workspace.

Privacy and Industry Implications

This development reduces the friction of AI personalization, making it significantly easier for professionals to generate content that maintains a consistent personal or brand voice. However, the automation of this process introduces substantial privacy and security considerations. To build an accurate stylistic profile, the AI must scan personal communications and private documents across multiple third-party applications.

This increases the surface area for the processing of sensitive personal information, as the AI analyzes not just the structure of the writing, but the content within those connected apps to determine stylistic nuances. The move signals a broader industry trend toward "agentic" personalization, where AI tools move from being passive recipients of instructions to active observers of user behavior.

What to Watch

As OpenAI continues testing the feature, it remains to be seen how the company will address the privacy concerns associated with scanning personal app data for stylistic training. Users and security researchers will likely look for granular controls that allow users to specify which apps—or which specific folders and threads—the AI is permitted to analyze. Whether this feature becomes a standard part of the ChatGPT interface or remains a limited experiment will depend on the balance between user convenience and data security.