OpenAI Tests Writing Style Feature So ChatGPT Mimics User's Tone

OpenAI has begun testing a feature meant to make ChatGPT write in a way that sounds like the user, not like a generic chatbot. Writing Style analyzes real messages, documents, and emails from connected apps to reproduce a specific person's tone, vocabulary, and sentence structure.

How the New Feature Works

The mechanism is simple in principle. The user connects selected apps, and ChatGPT scans their content for texts the person has previously written. Based on that, the model builds a stylistic profile covering word choice, sentence length, formality, and typical courtesy phrases.

The setup is split into three categories. Messages come from Slack, documents from Google Drive and Notion, and mail from Gmail. This lets the model distinguish how a given person writes a work email, how they chat with colleagues on Slack, and how they draft a report in Notion, applying the right register depending on the task's context.

ChatGPT will write in your voice by referencing examples from your connected apps - message displayed on the Writing Style feature's onboarding screen

Difference From Claude

Anthropic has long offered a similar solution through its Styles feature in Claude. There, the user has to manually upload a file or paste a text sample so the model can learn the preferred response style. OpenAI wants to go a step further and remove that burden from the user.

Instead of a single sample pasted into the chat window, ChatGPT would draw on the entire history of communication stored across connected services. In practice, that means a much larger and more varied set of examples than one manually uploaded document, which, according to the feature's description, should allow for a more faithful reproduction of style.

Testing Status and Availability

For now, Writing Style has only reached a small group of users as part of an experiment. OpenAI hasn't said when the feature might roll out more broadly, or whether it will make it into the production version of ChatGPT at all. The company regularly tests features on small groups, and some of them never reach all users.

The feature fits into a broader trend of integrating ChatGPT with everyday office tools through connectors to Gmail, Google Drive, Slack, and Notion, which OpenAI has been developing for some time. Writing Style uses the same connections, but instead of retrieving information, it uses them to build a stylistic profile of the user.

What It Means for Polish Users

For businesses and freelancers using ChatGPT to write emails, reports, or work messages, such a feature could mean less time spent correcting the tone of generated text. Many people in Poland already copy their own previously written passages into prompts to force the model into a more natural style, and Writing Style would aim to automate that step.

The solution also raises questions about the scope of access ChatGPT gains to a user's private and work correspondence. Connecting Gmail or Slack to a feature that analyzes writing style means the model processes the content of real messages, not just metadata, which in the context of company data may raise concerns among security teams.

There's currently no information on whether the data used to build the stylistic profile is isolated in any way from ChatGPT's other features, or how long it is retained. OpenAI has not published separate technical documentation on Writing Style, which makes it difficult to assess the actual scope of data being processed.