When you ask Copilot to generate a table, a list, or a structured report, it sometimes returns results in a layout you did not expect. This happens because Copilot interprets your prompt based on general patterns rather than your exact formatting requirements. The cause is that natural language prompts often lack explicit structural instructions, leaving Copilot to guess the output format. This article explains how to write prompts that force Copilot to follow a specific output format every time, using techniques such as structured prompts, delimiter markers, and format examples.
Key Takeaways: Controlling Copilot Output Format
- Structured prompt with explicit instructions: Tell Copilot the exact output structure, including column headers, row count, and delimiter type.
- Example output in the prompt: Provide a one- or two-line example of the desired format to anchor Copilot response.
- System message or context pane settings: In Copilot for Microsoft 365, use the system message field in the admin center to set default formatting rules for your tenant.
Why Copilot Does Not Always Follow Your Format
Copilot uses a large language model that generates text based on probability. When you ask a question without specifying the output format, the model selects the most likely structure based on its training data. For example, if you ask for a list of project tasks, Copilot might return a paragraph instead of a numbered list. The root cause is that natural language lacks structural markers. To get a predictable format, you must embed explicit formatting rules inside your prompt. This includes defining the output type, column names, row count, and any delimiters. In Copilot for Microsoft 365, you can also set tenant-wide formatting preferences through the admin center, which applies to all Copilot interactions.
Steps to Force Copilot to Follow a Specific Output Format
The following methods work in Copilot in Microsoft 365 apps such as Word, Excel, and Teams, as well as in Copilot on the web. Each method increases the likelihood that Copilot will return the exact structure you need.
Method 1: Use a Structured Prompt with Explicit Instructions
- Define the output type first
Begin your prompt with a clear statement of the format. For example, write “Create a table with three columns: Task, Owner, and Due Date.” Avoid vague phrases like “list the tasks.” - Specify the number of rows or items
Add a limit to prevent Copilot from generating too many or too few entries. Write “Include exactly five rows.” If you need dynamic data, say “Include one row per task from the current project.” - State the delimiter for inline formats
If you want a comma-separated list, say “Separate each item with a comma.” For a bullet list, say “Use a bullet point for each item.” This removes ambiguity. - Add a negative constraint
Tell Copilot what not to do. For example, “Do not use paragraphs or full sentences. Only output the table.” This prevents the model from adding explanatory text.
Method 2: Provide an Example Output in the Prompt
- Write a one-line example of the desired format
Insert a short sample after your instruction. For instance: “Example format: Task, Owner, Due Date. Task A, John, 2025-01-15.” This gives Copilot a concrete template to follow. - Use the example as a reference point
After the example, write “Follow the same format for all items.” This reinforces the pattern without additional explanation. - Test with a small request first
Ask Copilot to generate only two or three items using the example format. If the output matches, expand the request to the full data set.
Method 3: Set Default Formatting Rules in Copilot for Microsoft 365 Admin Center
- Open the Microsoft 365 admin center
Go to admin.microsoft.com and sign in with a Global Admin or Copilot Admin account. - Navigate to Copilot settings
Select Settings > Org settings > Copilot. This opens the Copilot configuration page. - Edit the system message
Under System message, enter a default formatting instruction. For example: “Always output tables with headers and use pipe delimiters.” This message is prepended to every Copilot prompt across your tenant. - Save and test
Click Save. Then ask Copilot a question in any app to verify that the default format is applied.
Common Mistakes and Limitations
Copilot Still Returns Paragraphs Instead of Tables
If Copilot ignores your table instruction, the prompt may lack a negative constraint. Add a sentence like “Do not write any sentences. Only output the table.” Also verify that you are using the correct app version. Copilot in Excel has stronger table support than Copilot in Word. For complex tables, use Copilot in Excel or create the table manually first, then ask Copilot to populate it.
Copilot Generates Extra Explanatory Text
Copilot often adds introductory or concluding sentences even when you ask for a raw output. To suppress this, place the format instruction at the very end of your prompt. For example: “List the top five risks. Output only the list, no explanation.” If the problem persists, use the system message method to set a tenant-wide rule that removes all explanatory text.
Copilot Does Not Follow the Example Format Exactly
When you provide an example, Copilot may mimic the content but change the structure. Ensure the example is minimal and uses the exact delimiter you want. If the example includes a header row, Copilot is more likely to repeat that header. Also avoid placing the example inside parentheses or quotes, as that can cause the model to treat it as a citation rather than a template.
Comparison of Prompt Techniques for Output Format Control
| Technique | Structured Prompt Only | Structured Prompt + Example |
|---|---|---|
| Description | Uses explicit instructions without a sample | Combines instructions with a concrete example |
| Success rate for tables | 60-70% | 85-95% |
| Success rate for lists | 70-80% | 90-98% |
| Requires negative constraint | Often needed | Less frequently needed |
| Best for | Simple formats like bullet lists | Complex formats like multi-column tables |
You can now force Copilot to return data in the exact format you need. Start by using structured prompts with explicit instructions and a short example. For tenant-wide control, set a system message in the Microsoft 365 admin center. For advanced users, combine the system message with inline examples for the highest precision. Test each prompt with a small sample before applying it to large data sets.