Imagine asking your AI assistant, “Which of our search campaigns lost the most conversions this month, and why?”, and getting an answer pulled from live ad data in seconds. That’s now a real option. In its August 2026 product roundup, published on 31 August, Microsoft Advertising confirmed that its Microsoft Advertising MCP server is available with read-only access through what Microsoft calls an “open flight” (an open pilot), letting AI assistants such as Microsoft 365 Copilot, ChatGPT and Claude run reports, campaign audits and keyword research on your account.
Microsoft isn’t alone. Meta and Google have also released official ways to connect AI assistants to their ad platforms. For small and mid-sized businesses, this is one of the most practical forms of AI automation to arrive this year. It also needs some sensible guardrails.
What is MCP, in plain English?
MCP stands for Model Context Protocol. It’s a standard way for AI assistants to connect to other software and use it as a tool. Instead of exporting a spreadsheet from your ad account and pasting it into ChatGPT, an MCP server lets the assistant fetch the data itself, with your permission.
What the major ad platforms now offer
Microsoft Advertising
According to Microsoft, its MCP server works with Microsoft 365 Copilot (via Copilot Studio), the Claude and ChatGPT desktop apps, and any MCP-compatible AI agent. It can explore campaigns, ad groups, ads and keywords, and pull metrics such as spend, clicks, impressions, CTR, average CPC, CPM, conversions, conversion rate, revenue and ROAS. Microsoft’s examples include pulling top products, search queries and month-over-month conversion changes, or asking for a campaign audit.
The connection itself is free, although you may need a paid subscription to your chosen AI assistant. Importantly, access is read-only for now, so the assistant can analyse but not change your campaigns.
Meta
Meta released its Ads AI Connectors, including an MCP server, in open beta on 29 April 2026. According to coverage from Passionfruit, the connector exposes 29 tools covering reporting, campaign management, catalogue management and signal diagnostics, and it can create as well as read. As a safeguard, anything an assistant creates lands in a paused state by default, and larger budget or targeting changes need human approval in Ads Manager.
Google Ads
Google maintains an open-source Google Ads MCP server on GitHub under its official googleads organisation. Its tools are focused on retrieving account information through the Google Ads API. Setting it up is more technical: it requires a Google Ads API developer token and Google Cloud credentials, so it’s better suited to businesses with a developer or an agency partner.
Why this matters for a small team
The biggest benefit isn’t flashy. It’s time. Pulling a weekly performance summary, spotting which keywords are wasting money, or comparing this month with last month are jobs that take a busy owner or marketer an hour or more. An assistant with direct access can do a first pass in minutes, across more than one platform.
That frees people up for the work AI is still bad at: understanding your customers, deciding what to offer, and judging whether a “winning” campaign is actually bringing in good customers.
The risks to manage
- Write access. Read-only tools can’t overspend. Tools that can create or change campaigns can. Start read-only, and keep approval steps switched on for anything that can spend money.
- Confident wrong answers. AI assistants can misread data or draw the wrong conclusion. Check important findings against the platform before acting.
- Who has access. Connecting an assistant to your ad account is like giving someone a login. Use accounts with the minimum permissions needed, and review connected apps regularly.
- Platform bias. Each platform’s data measures success its own way. Cross-check against your CRM or sales data.
What to do now
- Start with one read-only use case. A weekly “what changed and why” report is a great first job.
- Use a dedicated, limited-permission login for any AI connection to your ad accounts.
- Write your standard questions down. Consistent prompts give you consistent reports you can compare week to week.
- Verify before you act. Treat AI analysis as a first draft, especially for budget decisions.
- Only enable write access with safeguards. If you use a connector that can make changes, keep new items paused by default and require human approval.
- Connect the dots to revenue. The best automations combine ad data with CRM data, so you’re optimising for customers, not clicks.
AI assistants that can read your ad accounts are a genuinely useful step forward, as long as people stay in charge of the decisions. If you’d like help setting up AI-powered reporting or automation safely, the AI automation team at Big Digital is happy to talk it through.
Photo: Growtika on Unsplash