Workplace drafting often fails on tone rather than substance.
A message can be accurate and still sound unlike the person who is supposed to send it. That gap is familiar in email, Slack, and internal documents, where colleagues already know how someone usually writes.
OpenAI's ChatGPT Work now includes a writing style option designed to reduce that gap.
Instead of relying on a short prompt that describes tone, the feature looks at writing users have already produced in connected workplace tools such as Gmail, Google Drive, Slack, and SharePoint.
Those sources can provide a more useful picture of someone's writing than a one-line instruction because the real working voice already exists in email threads, chat replies, and documents.
OpenAI describes Work itself as the part of ChatGPT intended for longer, multi-step work and deliverables, rather than quick conversational questions. It can use connected workplace tools, files, and apps as context.
The Writing Style feature takes that idea a step further.
Instead of telling ChatGPT, "Make this sound like me," users can connect workplace sources and allow ChatGPT to examine their existing work writing for patterns in how they communicate.
Reports about the feature say those patterns can include vocabulary, sentence construction, recurring phrases, typical sign-offs, and smaller habits such as capitalization.
the goal is for a generated email to be closer to that style instead of producing a generic corporate version.
The practical value shows up in the editing step that usually comes after an AI draft.
Users often keep the content of an AI-generated message but rewrite the voice. The draft may be too polished, too generic, or too formal for Slack. It may use phrases they would never use, or add a sign-off that does not match their usual habit. A style layer that already knows recurring phrases, typical closings, and small patterns such as capitalization can shorten that rewrite pass.
That is most useful for people who send many short messages and recurring updates, because an off-tone sentence is easiest to notice there.
The setup is limited to ChatGPT Work rather than the regular chat experience. On the web, the option is under Settings, Personalization, and Writing style, and OpenAI says the setting then applies to Work messages on web and mobile.
It is available on paid plans that already include ChatGPT Work.
Users still have to connect the apps themselves. That is the tradeoff that makes the feature useful and also the reason to treat access with care. The model can only learn a writing style from material it is allowed to see. Connecting Gmail or SharePoint provides access to considerably more workplace material than providing a single document, so that connection is a work decision rather than something that needs to be enabled by default.
There is also a distinction between Writing Style and existing ChatGPT personalization features. Custom Instructions can tell ChatGPT what kind of writing people want, such as being concise or avoiding corporate jargon. A Writing Style feature instead tries to learn those characteristics from examples of how you already communicate.
The feature will not replace judgment.
Matching a sign-off is not the same as knowing which facts are current, which claims need a source, or which message should not be sent yet. A draft can sound exactly like its intended author and still contain an incorrect date, outdated information, or a claim that needs verification.
Used narrowly, the feature addresses a practical problem in workplace drafting: getting the first version closer to how users already write, so the time spent editing the draft can focus more on content and context rather than tone cleanup.
