Copilot Studio billing: is a “classic answer” counted per turn or per message?

A classic answer is billed per predefined response event, not under a documented “one credit per user turn” rule; however, Microsoft’s public documentation does not explicitly confirm whether four consecutive Message nodes in one topic execution produce four separately metered classic-answer events.

What can be confirmed

• A classic answer is a predefined, manually authored response and costs 1 Copilot Credit.
• One interaction can consume multiple feature charges simultaneously.
• Microsoft’s billing example calculates four classic answers as 4 × 1 credits, showing that multiple classic answers can be charged during a run rather than being collapsed automatically into one charge per user turn. ^1^

For your sequence:

1. “Please wait a moment.”
2. “Checking the market research…”
3. “Writing the draft…”
4. Final answer

Do not budget on the assumption that the entire sequence is necessarily one credit. The documented unit is the classic-answer event, but the documentation does not map that event explicitly to each Message node or each outgoing Teams activity.

If the intermediate status messages are not operationally necessary, combining them into one Message node is the conservative design for avoiding potentially separate classic-answer events.

Adaptive Cards

An Adaptive Card can be part of a message activity: Copilot Studio describes message activities as carrying text, media, or Adaptive Card content. ^2^

For a noninteractive card, Microsoft specifically supports adding the card to a Message node:

1. Open the Message node.
2. Select Add → Adaptive card.
3. Select Edit adaptive card, enter the payload, and save it. ^3^

The billing documentation does not state that:

• a card within a Message node incurs an additional classic-answer credit beyond the surrounding response, or
• text and a card in the same Message node are billed as two classic answers.

Therefore, a separate card charge cannot be confirmed from the published rules. Treat text plus a card in one Message node as one response design, but validate actual metering before using that assumption for a cost forecast. An interactive Adaptive Card node is also a distinct conversation node, yet Microsoft does not publish a card-specific Copilot Credit rate. ^4^

How to inspect consumption

Copilot Studio: agent-level totals

Open the agent and go to Monitor:

• Overview shows total billed Copilot Credits for the selected period.
• See billing opens more detail.
• Billing trend shows credits over time.
• Cost distribution breaks consumption down by activity type, including classic answers.

The data can take several hours to appear. ^5^

This view is useful for comparing controlled test runs, but it does not provide a node-by-node ledger.

Power Platform admin center: detailed reporting

An administrator can use:

1. Power Platform admin center → Licensing.
2. Under Products, select Copilot Studio.
3. Select Environments and open the relevant environment.
4. On Copilot Credits capacity, find Copilot credit consumption details.
5. Select Download, choose the timeframe, and select Start.

The detailed report includes:

• Agent name
• Product/AI feature
• Channel
• Model
• Knowledge source/tool
• Billed and non-billed credits

It does not report usage for an individual conversation or message, so it cannot directly identify “Message node 2 used one credit.” ^6^

Practical validation

Because embedded test chat activity is free and excluded from Copilot Credit reports, perform the comparison through the published Teams agent, not the Copilot Studio test pane. ^7^

Use two temporary topics during a quiet reporting period:

• Test A: one Message node containing all four lines.
• Test B: four consecutive Message nodes containing one line each.

Run each test a known number of times with the same account and no actions, flows, grounding, or generative answers. After reporting updates, compare Classic answer consumption for that agent in the downloaded environment report. This is the available way to determine the effective node/activity granularity in your tenant and current service version; the reports cannot expose individual messages directly.

If the result remains ambiguous because other traffic is present, open a Microsoft support request for an authoritative metering trace.

References

1. Billing rates and management
2. Send an event or activity
3. Send a message
4. Ask questions with Adaptive Cards
5. View agent's billing consumption
6. Manage Copilot Credits and capacity for Copilot Studio
7. FAQ for Copilot Studio billing and licensing