On August 31, 2026, Microsoft Advertising announced in its official product monthly report that its MCP server is open read-only access via open flight. Advertising teams can connect real-time Microsoft Advertising data to AI assistants like Microsoft 365 Copilot, ChatGPT, and Claude, running reports, account audits, and keyword research in conversations without repeatedly exporting spreadsheets, copying data, or switching backends.
The key to this update isn't adding another chat entrance, but that delivery data enters the team's existing AI workflow for the first time in a standardized context. Readers unfamiliar with MCP can first check out the role and boundaries of the model context protocol.
Read-only design first addresses "understanding" and does not replace execution for others
Microsoft's case shows that Stagwell compressed campaign audits from hours to minutes; Conversios used it to check activity health, UET conversion goals, Merchant Center readiness, and audience signals for over 120 accounts; Realtor.com sent competitive keyword data into the Claude workflow while retaining manual review before launch.
"Read only" is very important. Currently, MCP servers can read and analyze data, but cannot directly modify budgets, bids, or delivery status in the assistant. This reduces the risk of misoperation and makes it easier for teams to first establish audit rules and verify report calibers before considering whether to enable write capabilities in the future. When multiple customer accounts are involved, connections should still be made with least privilege, and records who made the query, what suggestions the model generated, and who ultimately approves it.
The threshold for ad optimization has shifted from moving data to defining criteria
The same monthly report also confirmed that AI Max search is officially available, helping advertisers understand intent and adapt to new queries; Account links have added Dual-ID verification, further tightening authorization. Looking at these updates together, Microsoft is integrating AI optimization, data access, and permission verification into a unified operational chain.
The most directly affected are agencies, cross-account delivery teams, and e-commerce operations. In the future, efficiency gaps will no longer mainly come from who exports the list faster, but from who can write "abnormal spending, conversion gaps, and keyword opportunities" into clear and reusable audit standards. MCP shortens the distance from data to recommendations, but whether recommendations are reliable still depends on conversion tracking quality, account boundaries, and manual review.
