ChatGPT for Word Launches: Access, Cost, Privacy, and Admin Controls

ChatGPT for Word is now available as a sidebar inside Microsoft Word. It can work with the open document and selected text to draft, summarize, reorganize, revise, and adjust basic formatting. OpenAI says the add-in is available across ChatGPT plans, while workspace and Microsoft 365 administrators control organizational access. Starting October 1, 2026, OpenAI plans to enable Word access by default, so administrators should review settings before that date.

The same ChatGPT Microsoft add-in used for Excel and PowerPoint now supports Word. A user signs in with a ChatGPT account, selects the appropriate workspace, and can ask the sidebar to work on the open document. OpenAI's documentation says the add-in cannot reference arbitrary local files; relevant material from another file must be pasted or supplied through a supported connected app.

The product is not the same as Microsoft 365 Copilot. It uses the user's ChatGPT account and OpenAI workspace controls, even though Microsoft 365 administrators must also permit the add-in.

OpenAI says Word uses token-based pricing tied to the selected model. Document length, requested work, cached input, and generated output affect usage. On plans with shared premium allowances, Word can consume the same pool used by Codex and other eligible features. A long-document rollout therefore needs usage monitoring and spend limits, not only a seat count.

Published availability does not establish a predictable cost per memo or contract. Teams should measure tokens, retries, review time, and accepted output on their own document mix.

The add-in can process the open document, which may contain confidential, personal, privileged, or regulated information. OpenAI states that Business, Enterprise, and Edu inputs and outputs are not used to train its models by default. That statement does not replace an organization's classification rules, retention review, connected-app permissions, contractual requirements, or Microsoft 365 add-in governance.

Conversations in the add-in are separate from ordinary ChatGPT history, and OpenAI says memory and skills do not carry over. Supported apps can add context from services such as Outlook or SharePoint when the plan, workspace, and source permissions allow it; plugins are not currently supported in Word.

The launch does not prove that edits preserve every fact, citation, defined term, table, chart, style, or legal meaning. OpenAI explicitly warns that complex formatting can need manual adjustment and that ChatGPT can change content users intended to keep. It also does not establish net productivity once correction, approval, usage, and governance costs are included.

Use 30 permission-safe documents spanning short memos, long reports, tables, citations, defined terms, sensitive classifications, and complex formatting. Compare the original and revised files. Record factual changes, omitted qualifications, broken references, formatting damage, token use, reviewer time, and final acceptance. Keep version history, require approval before external sharing, and test disabling access through both OpenAI and Microsoft 365 administration.

Workspace and Microsoft 365 administrators should inventory eligible users, review default-on behavior before October 1, define prohibited document classes, set spend controls, confirm connected-app permissions, and publish a review checklist. Individual users should save a copy before substantial edits and verify facts, numbers, citations, defined terms, and formatting before sharing.

This is a research-based news analysis using OpenAI's current product, release, privacy, and administrative documentation. DiscoverAI has not audited the add-in or measured it across a long-term enterprise deployment. Features, defaults, limits, pricing, and policies can change.