When you ask Copilot a question in Microsoft 365 and it replies only with the message “Content filtered” and no actual output, you cannot see the answer you requested. This problem typically happens because Copilot’s built-in content moderation system flagged your prompt or the generated response as violating a policy rule, and it blocked the result without displaying an explanation. The root cause can be an overly strict tenant-level policy, a misconfigured sensitivity label, or a prompt that triggers a false positive in the safety filter. This article explains why the filter blocks content, walks through the exact steps to identify and resolve the cause, and covers related failure patterns you may encounter.
Key Takeaways: Fix ‘Content Filtered’ in Copilot
- Microsoft 365 admin center > Copilot > Content filters: Adjust the sensitivity level or turn off specific filter categories for testing.
- Copilot pane > Settings > Data protection: Check if data loss prevention policies are blocking specific content types.
- Prompt rewrite technique: Remove ambiguous terms, numbers, or phrases that may trigger false positives in the safety classifier.
Why Copilot Shows ‘Content Filtered’ Instead of Output
Copilot uses a multi-layered content moderation system that examines both the user prompt and the generated response. The system applies Microsoft’s Responsible AI policies, which include rules against harmful, offensive, or restricted content. When the classifier detects a potential violation, it blocks the output and returns the generic “Content filtered” message. This behavior is by design to prevent inappropriate content from being displayed.
The filter operates on three levels:
1. Prompt-level filtering
Before Copilot processes your question, the system checks the text against a list of prohibited categories such as hate speech, violence, self-harm, and sexual content. If the prompt matches any category, the filter blocks the request immediately.
2. Response-level filtering
After Copilot generates an answer, the same classifier scans the output. If the response contains text that the model considers borderline or risky, the filter removes it and shows the “Content filtered” message.
3. Tenant-level policy enforcement
Microsoft 365 administrators can configure custom policies in the Microsoft 365 admin center that extend or tighten the default filter behavior. These policies include data loss prevention rules, sensitivity labels, and conditional access policies that may cause false positives for legitimate business queries.
Steps to Diagnose and Fix the ‘Content Filtered’ Issue
Follow these steps in order to identify the exact cause and resolve the problem.
- Reproduce the issue with a known safe prompt
Open Copilot in Microsoft 365 and type a simple, neutral question such as “What is the capital of France?” If this prompt also returns “Content filtered,” the problem is likely a tenant-level policy or a global filter setting. If the safe prompt works, the issue is specific to your original query. - Check Microsoft 365 admin center for content filter settings
Sign in to the Microsoft 365 admin center. Go to Settings > Org settings > Copilot > Content filters. Review the current sensitivity level. If it is set to “High,” lower it to “Medium” temporarily and test again. If the filter is blocking specific categories, consider disabling the category that matches your prompt type. - Review data loss prevention policies
In the Microsoft Purview compliance portal, navigate to Data loss prevention > Policies. Look for any policy that applies to Copilot or Microsoft 365 Copilot. Check if the policy blocks content containing financial data, personal information, or confidential terms. If a policy is too broad, edit the policy rule to exclude legitimate business queries or add the affected user group to an exception list. - Test with a different user account
Ask a colleague with the same license to run the same prompt. If the colleague gets a normal response, the issue may be tied to your account’s sensitivity label or conditional access policy. Check your account’s assigned sensitivity label in Azure AD. If the label has a classification such as “Highly Confidential,” it may trigger stricter filtering. - Rewrite the prompt to avoid ambiguous terms
Replace words that could be interpreted as harmful or sensitive. For example, change “terminate employee contracts” to “end employee contracts.” Remove numbers that look like social security numbers or credit card patterns. Use clear, neutral language and avoid slang or idioms. - Clear the Copilot session cache
In Microsoft Edge or Google Chrome, open the browser’s developer tools by pressing F12. Go to the Application tab, select Storage, and click Clear site data. Reload the Copilot pane and try the prompt again. This removes any cached session state that may interfere with the filter. - Contact Microsoft support with filter logs
If none of the above steps work, gather the following information: the exact prompt that failed, the timestamp, your tenant ID, and the user account UPN. Submit a support request through the Microsoft 365 admin center. Ask the support team to review the content filter logs for your session to identify the specific rule that blocked the output.
If Copilot Still Has Issues After the Main Fix
Copilot blocks output for all users in the tenant
If every user in your organization receives “Content filtered” for every prompt, the issue is likely a global tenant policy. Go to the Microsoft 365 admin center and check Settings > Org settings > Copilot > Content filters. Ensure that the filter is not set to “Block all content.” Also verify that no data loss prevention policy is applied to the entire tenant without exceptions.
Copilot blocks output only for specific file types
When you ask Copilot to summarize a Word document or an Excel spreadsheet and the response is filtered, the file itself may contain content that triggers the safety classifier. Open the file and scan it for any text that could be flagged, such as confidential disclaimers, profanity, or financial figures. Remove the offending content or save a copy without that text and try again.
Copilot returns ‘Content filtered’ after a recent update
Microsoft occasionally updates the content moderation model, which can cause new false positives. If the issue started after a known update, check the Microsoft 365 Message Center for announcements about changes to Copilot content filtering. You may need to wait for a model refresh or adjust your tenant policies to accommodate the new rules.
Copilot Content Filter Levels: Comparison
| Item | Low Sensitivity | High Sensitivity |
|---|---|---|
| Description | Blocks only explicitly harmful content such as hate speech and violence | Blocks content that may be ambiguous or borderline, including financial data and personal information |
| False positive rate | Low | High |
| Recommended for | General business queries with no confidential data | Tenants handling sensitive data or with strict compliance requirements |
| Admin control | Can enable or disable specific categories | Can enable or disable specific categories, plus set data loss prevention rules |
Now you understand why Copilot shows “Content filtered” and have a clear set of steps to diagnose and fix the issue. Start by testing with a simple prompt and checking the admin center content filter settings. If the problem persists, rewrite your prompt with neutral language and clear your browser cache. For persistent tenant-wide issues, contact Microsoft support with the filter logs to identify the exact blocking rule.