You are working on a lengthy Word document and Copilot suddenly stops responding. The Copilot pane shows a spinning icon or a message that the request timed out. This happens because Copilot has a hard limit on how much text it can process in a single session. This article explains the technical timeout limits, how they affect long documents, and what you can do to work around the issue.
Key Takeaways: Copilot Timeouts in Long Word Documents
- Copilot context window limit of 4096 tokens: The model can only read and respond to roughly 3000-4000 words at a time, causing timeouts on documents exceeding that length.
- Select specific paragraphs before prompting: Highlight only the relevant section of a long document to stay within the token limit and avoid timeouts.
- Break the document into smaller sections: Use Word headings or separate files to reduce the amount of text Copilot must analyze in one request.
Why Copilot Time Out Occurs on Long Documents
Copilot in Word uses a large language model that processes text in chunks called tokens. A token is roughly four characters in English, so 100 tokens equal about 75 words. The model has a fixed context window, currently set at 4096 tokens. This includes the user prompt and the document text Copilot reads. When the document exceeds this limit, Copilot cannot analyze the full content and the request times out.
The timeout is not a network or server error. It is a design constraint of the underlying AI model. Microsoft sets the limit to balance response speed and accuracy. Processing more tokens takes longer and can produce less coherent results. The timeout protects the user experience by stopping requests that would fail or return poor-quality responses.
Token Limit in Practice
A 10-page Word document with standard formatting contains roughly 5000 words or 6700 tokens. This exceeds the 4096 token limit by over 60 percent. Copilot will time out on the entire document. Even a 5-page document with dense tables or bullet points can hit the limit. The model also counts the prompt itself, so a long instruction reduces the space available for document text.
How Copilot Reads Document Content
Copilot does not read the entire document file. It reads the text visible in the current selection or the active paragraph range. When you ask a general question like “Summarize this document” without selecting text, Copilot tries to read the whole document. This triggers the timeout. If you select a single paragraph, Copilot reads only that paragraph and responds quickly.
Steps to Avoid Copilot Timeout on Long Documents
Follow these steps to keep Copilot responsive when working with long Word documents.
- Select a specific section before prompting
Highlight the paragraph or section you want Copilot to analyze. For example, select the introduction of a 20-page report. Then type your prompt. Copilot reads only the selected text, staying within the token limit. - Use Word headings to break the document
Apply heading styles like Heading 1, Heading 2, and Heading 3 to your document sections. Copilot uses headings to identify logical breaks. You can then ask Copilot to work on a specific heading: “Summarize the section under Heading 2.” This limits the text Copilot processes. - Split the document into smaller files
Divide a long document into multiple files of 3-5 pages each. Store them in the same SharePoint folder or OneDrive folder. Copilot can reference each file separately. This method works best for reports, manuals, or proposals with distinct chapters. - Reduce the prompt length
Keep your prompts under 50 words. Long prompts consume tokens that could otherwise be used for document text. For example, instead of “Can you please provide a detailed summary of the key points in the third paragraph of the financial analysis section,” write “Summarize the third paragraph.” - Use the Copilot compose feature for short text
If you need Copilot to generate new content rather than analyze existing text, use the Compose button in the Copilot pane. This feature sends only your prompt to the model, not the full document. It avoids timeout entirely for writing tasks. - Restart Copilot after a timeout
Close the Copilot pane and reopen it. This clears the cached context. Then select a smaller portion of the document and retry your prompt. A fresh session often resolves transient timeout issues.
If Copilot Still Times Out After These Steps
Some situations require additional adjustments. Here are common scenarios and their fixes.
Copilot Times Out on a Single Paragraph That Is Too Long
A paragraph with more than 4000 words will still time out even when selected. Split the paragraph into two or three shorter paragraphs. Each paragraph should stay under 3000 words. Then select one paragraph at a time.
Copilot Returns an Error Message Instead of a Timeout
If you see “Something went wrong” or “We couldn’t complete your request,” the issue may be a server-side problem. Check the Microsoft 365 Service Health dashboard for Copilot service status. If the service is healthy, reduce the document selection further and retry.
Copilot Works on Short Documents But Not Long Ones
This confirms the token limit is the cause. No amount of retrying will make Copilot process a document that exceeds 4096 tokens. You must reduce the text Copilot reads. Use the selection method described above.
Copilot Free vs Copilot Pro vs Copilot for Microsoft 365: Token Limits
| Item | Copilot Free | Copilot Pro | Copilot for Microsoft 365 |
|---|---|---|---|
| Context window | 4096 tokens | 4096 tokens | 4096 tokens |
| Document reading in Word | Not available | Not available | Yes, with 4096 token limit |
| Priority access during peak | No | Yes | Yes |
| Best for long documents | No | No | Yes, with selection workaround |
Copilot Free and Copilot Pro do not include the Word document integration feature. Only Copilot for Microsoft 365 can read Word documents. All three tiers share the same 4096 token context window for chat. The timeout issue affects only Copilot for Microsoft 365 when reading long documents.
You can now work with long Word documents without Copilot timing out. Select specific sections, use headings, or split files to stay within the token limit. For very long paragraphs, break them into smaller pieces. This approach keeps Copilot responsive for summaries, rewrites, and Q&A on any document length.