Copilot Narrative visual not loading in Power BI Service; "Content Unavailable"

Hi acfeigen,

Thank you for reaching out to Microsoft Fabric Community.

Thank you ShivekMaharaj and DataTako for the prompt response.

As we haven’t heard back from you, we wanted to kindly follow up to check if the solution provided by the user's for the issue worked?  or let us know if you need any further assistance.

Thanks and regards,

Anjan Kumar Chippa

Hi @acfeigen,

We wanted to kindly follow up to check if the response provided by the user's for the issue worked?  or have you raised a support ticket?

Create a Fabric and Power BI Support Ticket - Power BI | Microsoft Learn

Thanks and regards,

Anjan Kumar Chippa

Hi acfeigen​,

An F2 capacity should be sufficient. Microsoft’s Copilot enablement guidance for Power BI currently supports paid Fabric capacities from F2 upward, and the Copilot narrative requirements do not specify a higher minimum SKU for this visual.

One important difference is that the published narrative is not simply static text generated in Desktop. Power BI stores the prompt with the report metadata and generates the summary again when the report loads. That would explain why Desktop can succeed while the Service displays Content unavailable.

Since you have already reproduced it with a minimal report, I would check these Service-specific items next:

• Confirm that the published report and semantic model are both in the intended paid F2 workspace.
• Verify the F2 capacity’s region against Microsoft’s supported Fabric region availability.
• If the capacity is outside the United States or EU data boundary, confirm that Data sent to Azure OpenAI can be processed outside your capacity’s geographic region, compliance boundary, or national cloud instance is enabled.
• Check the capacity’s delegated tenant settings. A capacity administrator can override the tenant-level Copilot setting, so both can appear enabled while the specific capacity behaves differently.
• Confirm that the workspace is not using Private Link or a closed-network configuration, because Microsoft currently lists those environments as unsupported for Copilot.
• Create the same narrative directly in the Power BI Service rather than publishing the Desktop-created one. If that also fails, it further isolates the issue to the capacity or Service configuration.

If your account is assigned to a separate Fabric Copilot capacity, I would also confirm that it is registered in the tenant’s home region. Microsoft explains that a Fabric Copilot capacity can bill Copilot usage for Power BI workspaces, including those on smaller Fabric SKUs, but it must reside in the tenant’s home region.

Given that the issue persists in a new workspace with a one-table test model, I would not spend much more time changing the report. If the region, cross-geo setting and delegated capacity settings are all correct, this looks suitable for a Microsoft support case. I would include the workspace and capacity IDs, report ID, capacity region, approximate failure time and any request or activity ID available from the Service.

I hope this helps narrow it down. The important point is that F2 itself is supported, so the failure is more likely related to the Service-side capacity configuration, regional processing requirements or a product issue than the size of the SKU.

AI-assisted drafting: AI was used to help structure and phrase this response. I reviewed and validated the technical content before posting.

The clue is that it generates fine in Desktop but fails in the Service, because those two use different paths to the AI backend. Desktop can reach it while your capacity can't, which almost always points at the Azure OpenAI region availability for your capacity rather than the F2 size. Copilot works on F2 and up now, so capacity size isn't your problem.

What usually causes exactly this "spins then Content Unavailable": your capacity's region doesn't have Azure OpenAI available, so the Service can't complete the call. The fix is the tenant setting "Data sent to Azure OpenAI can be processed outside your capacity's geographic region" (in the Admin portal, near the Copilot setting). Enable that and republish, and the narrative should render.

If that setting is already on, double-check the capacity's actual region against the list of regions where Copilot/AOAI is supported, since a capacity in an unsupported region will fail even with everything else correct.

If enabling the cross-region setting fixes it, please mark this as the solution so others on smaller capacities find it faster.

The problem still subsists.

I've tried creating a new Fabric workspace, created a dummy dataset in the service and created a new report with 1 regular visual and 1 narrative visual and it still errors out.

Workspace Settings:

Tenant Settings:

Copilot Capacty: Enabled

Data sent to Azure OpenAI can be processed outside your capacity's geographic region... Enabled

Admin portal:

Users can use Copilot and other features powered by Azure OpenAI: Enabled for all users in capacity

Semantic model:

Approved for Copilot: enabled