OpenAI is testing a new ChatGPT feature called Writing Style that could allow the AI assistant to learn how users naturally communicate by referencing content from their connected applications.

The experimental feature is currently available to a small group of users and can use examples from three broad areas: messaging, documents and email.

Apps shown as supported during testing include Slack for messaging, Google Drive and Notion for documents, and Gmail for email.

Instead of repeatedly telling ChatGPT to make an email sound more professional, casual or similar to previous writing, users could allow the system to reference existing examples of how they normally communicate.

This could help ChatGPT recognise differences in how the same person writes depending on the context. A quick message to a colleague, for example, may have a very different style from a formal email or a long business document.

The approach differs slightly from existing AI writing personalisation tools. Anthropic's Claude, for example, has offered Styles, which allows users to provide writing samples that can be used to create a custom writing style.

OpenAI's experiment could go further by allowing ChatGPT to reference writing that already exists within connected services rather than requiring users to manually provide examples.

OpenAI headquarters, Image: Envato
OpenAI headquarters, Image: Envato

Such a feature could make AI-generated writing feel considerably more personal. Instead of producing the same generic tone for different users, ChatGPT could potentially learn vocabulary, sentence structure and communication patterns from real examples.

It could also make the feature useful for professionals who regularly use ChatGPT to draft emails, documents and workplace messages.

OpenAI has confirmed that Writing Style is being tested, but the company has not announced when, or whether, the feature will become widely available.

For now, its limited rollout provides another indication of where AI assistants are heading. The next stage of personalisation may not simply involve AI remembering what users prefer, but learning how they communicate across the digital tools they already use.