Copilot Studio Knowledge Sources: Choose and Check Them

Copilot Studio agents answer questions by searching the knowledge sources you connect to them. The quality of the answers depends directly on the quality and currency of those sources. Before your agent can be useful in production, you need to choose the right sources, confirm that the agent has access to them, and test that it is actually finding and using the content you expect.

A tool like fenbs keeps the reviewed project context and the decisions behind your source choices in a form that any connected AI can read, independent of the agent platform itself. That matters when the agent configuration changes or when you are working across more than one AI tool.

What knowledge sources Copilot Studio supports

Copilot Studio supports several kinds of knowledge sources. The most common are SharePoint sites and document libraries, public websites, files uploaded directly, and Azure AI Search indexes for custom or enterprise knowledge bases. Each type has different setup requirements and different behavior in terms of how current the indexed content is.

Check Microsoft's Copilot Studio documentation↗ (opens in a new tab) for the current list of supported source types and any feature availability restrictions by region or plan. The source types available change as the platform evolves.

Choosing the right source type for your agent

The best source type depends on where your content lives and how often it changes. SharePoint works well for internal documentation, procedures, and policies that are maintained by your team in Microsoft 365. Public website sources work for information your agent should retrieve from public pages. Azure AI Search is the right choice when you need to index large volumes of structured enterprise data or non-SharePoint content.

• SharePoint: good for internal documentation and policies. Content is indexed from SharePoint sites the agent has permission to read.
• Uploaded files: quick to set up but static. Updating the content requires re-uploading the files.
• Public websites: the agent indexes specified URLs. Check the freshness settings to understand how often the index is refreshed.
• Azure AI Search: for large or structured enterprise knowledge bases. Requires an Azure subscription and index configuration.
• Custom APIs: developers can add custom knowledge sources through the extensibility model.

Checking SharePoint knowledge source access

A SharePoint knowledge source requires that the Copilot Studio agent has read access to the SharePoint sites you specify. In Microsoft 365, this typically means the agent's service account or the user running the agent must have at least read permission on the SharePoint site or document library.

Common access issues: the SharePoint site uses restricted access groups that do not include the agent, the document library has unique permissions that override site inheritance, or the M365 admin has blocked third-party app access to SharePoint content. Check with your M365 admin if the agent returns no results from a SharePoint source you have connected.

For the status of a knowledge source that shows as "in progress" in Copilot Studio, this typically means the source is still being indexed. Large SharePoint sites can take time to index fully after you add them. Refresh the page and check again after a few minutes.

Testing whether your agent uses the knowledge source

Always test the agent after adding a knowledge source. Ask a question you know the answer to from the connected source. If the agent answers correctly and cites the source, the knowledge source is working. If it does not find the right answer, check whether the content is actually indexed and whether the query you used would match the way the content is written.

Also test a question the source does not answer. The agent should say it does not know rather than generating a plausible-sounding answer from its own training. If it is generating answers not grounded in your sources, check the generative AI settings for the agent and consider whether to restrict it to only answer from connected sources.

Freshness and maintenance

Knowledge sources become stale. A SharePoint page updated last month reflects last month's process. A website source indexed on a fixed schedule may not have the current version of a page. Build a review process into how you maintain the agent: check key source documents on a schedule and update or re-index when the content changes significantly.

Record the decisions behind your source choices so you can review them later. Why was this SharePoint site included and not that one? Which content categories are approved for this agent to answer? Those decisions are easy to lose track of as the agent evolves.

How to do this with fenbs

1. Create a free account and add a project for the Copilot Studio agent.
2. Record the approved knowledge sources as Decisions and rules: which sources are included, what they cover, who approved them, and the review schedule.
3. Add tasks for the setup, testing, and first production review of the agent.
4. When a source is updated or replaced, record the change as a new decision that supersedes the previous one.
5. Connect a supported AI assistant over MCP from Integrations to keep the broader project context available alongside the agent configuration.

For related topics, see Copilot Studio for a general overview, Microsoft Copilot agents for the broader agent model, and Copilot instructions examples for how to write effective instructions for an agent.