ChatGPT for Word goes default-on for Business, Enterprise and Edu from 1 October 2026
ChatGPT's Word add-in becomes enabled by default on 1 October 2026. Admins must act now if they want to keep it off.
OpenAI confirmed on 17 September 2026 that ChatGPT for Word is generally available across all plan tiers, including Free. For organisations running ChatGPT Business, Enterprise, or Edu workspaces, there is a specific date to note: from 1 October 2026, the Word add-in will be enabled by default. If you want it off, you need to act before that date.
What the add-in actually does
ChatGPT appears in Word as a sidebar, accessible from its own button on the Word ribbon under Add-ins. From there you can draft content from rough notes, summarise an existing document, revise selected text, and adjust headings and formatting. It is the same add-in that already brings ChatGPT into Excel and PowerPoint, so Word completing that trio means organisations already familiar with it in those applications will find the behaviour consistent.
To get started, install the ChatGPT add-in from Microsoft AppSource, open it from the Word ribbon, and sign in with a ChatGPT account.
OpenAI is candid about the limitations. Complex formatting, tables, and charts may need manual fixes after ChatGPT makes changes, and the company notes that ChatGPT “can make mistakes and change content you intended to keep.” Saving a copy before asking it to make substantial revisions is sensible advice.
The 1 October 2026 date is the one that matters for admins
For Business, Enterprise, and Edu workspaces, access to ChatGPT for Word can be managed through the ChatGPT admin console. Admins can turn it on or off independently of other features. Right now, that toggle defaults to off. On 1 October 2026, it flips to on automatically.
If your organisation has not yet decided whether it wants ChatGPT available inside Word documents, that deadline gives you a concrete window to work with. Log into the ChatGPT admin console before 1 October 2026, and disable the feature if needed. Once that date passes, it will be active for all eligible users in your workspace unless you have already turned it off.
There is a second control layer that sits entirely outside OpenAI’s console: Microsoft 365 admins must separately allow the add-in at the tenant level before it can appear for employees. Both switches need to be in the right position. OpenAI enabling it by default does nothing for users if the Microsoft 365 side has not approved the add-in, and vice versa. Coordinate between whoever manages your OpenAI workspace and whoever manages your Microsoft 365 tenant.
How it is priced, and why that matters
This is where Business, Enterprise, and Edu customers need to read carefully. Word usage is not included as a flat feature within your subscription. It is billed at token-based API rates for whichever model is used, and it draws from the same shared usage pool as Codex, ChatGPT Work, Excel, PowerPoint, and Workspace Agents.
The default model where available is GPT-5.6 Sol, priced at $5.00 per million input tokens and $30.00 per million output tokens. Other models in the family are cheaper: Terra at $2.00 input and $12.00 output, and Luna at $0.20 input and $1.20 output. Workspace admins can control which models are available to their users.
What this means in practice: a user who asks ChatGPT to summarise a long report is consuming tokens proportional to the document’s length, their instructions, and the generated response. It is not a fixed cost per task. Organisations that allow heavy document work through the add-in should monitor usage through the admin console, particularly if multiple teams are using it alongside Codex or other agentic features that draw from the same pool. Full rate card details are available in OpenAI’s Business and Enterprise/Edu pricing documentation.
On data handling: inputs and outputs in Business, Enterprise, and Edu workspaces are not used to train OpenAI’s models by default. Consumer plan users are subject to their own data controls.
A free preview period for Business and Enterprise
From 17 September 2026 through 30 September 2026, ChatGPT Business and Enterprise customers can use GPT-5.6 Sol in Word without it counting against token allowances. That preview applies only to Word, not to the other Office integrations or agentic features. It is a useful window to test how your teams actually use the feature before the default-on date arrives and token consumption begins in earnest.
What this means for your organisation
The practical checklist is short but important:
- Before 1 October 2026, decide whether you want ChatGPT for Word enabled in your workspace. If not, disable it in the ChatGPT admin console now.
- Coordinate with your Microsoft 365 admin to confirm whether the ChatGPT add-in is permitted at the tenant level. This is a separate decision from the OpenAI console setting.
- Review your token usage in the admin console, particularly if your teams already use Codex or other agentic features. Word draws from the same shared pool, so unexpected document activity could affect other workflows.
- Use the free preview period (to 30 September 2026) to run a pilot with Business or Enterprise users before costs begin.
For organisations that have been cautious about AI in documents specifically, the auto-enable date is the most important thing here. The feature will not appear silently if your Microsoft 365 admin has not approved the add-in, but if both sides are already permissive, it will switch on without any further action from you.