Claude for Google Ads: build a Google Shopping AI agent

Yes, Claude can work with Google Ads: you connect it to your account through an MCP server, a connector that lets the assistant read your campaign data and answer questions about it in plain language. ChatGPT works the same way. For e-commerce, the useful setup reads Google Ads and Merchant Center together, because most questions are about products, not campaigns. The Clarmix MCP server does that for Google Shopping, adds Clarmix's product labels and insights, and retrieves up to 90 days of data.

This guide is for PPC consultants who manage Google Ads for online stores and want an AI agent that helps with day-to-day decisions. It covers what MCP is, the ways to connect Google Ads to Claude or ChatGPT, what a Google Shopping agent can do, and how to set one up.

Can Claude work with Google Ads?

Claude and ChatGPT don't see your ad accounts by default. They need a connector. Most connectors are built on the Model Context Protocol (MCP), an open-source standard for connecting AI applications such as Claude or ChatGPT to data sources, tools and workflows (Model Context Protocol).

In Claude, connectors let it access your apps, retrieve data and take actions in connected services, with the same permissions you have in that service. Available connectors are listed in Claude's Connectors Directory (Claude Help Center). In ChatGPT, connected apps do the same job: you choose an app, connect an account if prompted, and tell ChatGPT what you want (OpenAI Help Center).

Once Google Ads is connected, an agent is simply the assistant plus that access: you ask a question, it fetches the data it needs and answers with your numbers.

What are the ways to connect Google Ads to an AI assistant?

As of October 2026, these are the main routes for an e-commerce account:

Sources for the Google options: Ask Advisor, the Google Ads MCP server guide and the Merchant API MCP Access Service. Many third-party connectors also exist; check on each vendor's site whether it can make changes and which data it covers.

Why does an e-commerce agent need Merchant Center data too?

For an online store, the unit of decision is the product. A drop in Shopping spend can come from a bidding change, but just as often from products disapproved in Merchant Center, a price mismatch or a feed that stopped updating. An agent that only reads Google Ads will describe the drop; one that also reads Merchant Center can tell you which products caused it. Account-level issues matter even more: if Merchant Center flags a policy problem, every product in the account can stop serving, as explained in our guide to Merchant Center account suspensions.

The same goes for structure. Which products deserve their own asset group, which ones spend without converting, which ones Google barely shows: each answer needs ad performance and product data side by side. If you need a refresher on the Merchant Center side, start with our guide to Merchant Center account setup and key settings.

What can a Google Shopping AI agent do for a PPC consultant?

Here are three questions PPC consultants ask the Clarmix MCP server, and what the agent needs to answer them.

"Tell me how my campaigns have evolved these last 3 months"

The agent pulls 90 days of campaign data and summarizes the trend: spend, conversion value, ROAS, and which campaigns moved most. It can then go one level down, into the products and asset groups behind the change. That turns a monthly client review from an export job into a conversation you check and refine.

"What are my costly products I need to remove?"

This is where product data matters. Clarmix labels each SKU daily as Profitable, High Potential, Low Potential, Costly, Zero Conversion or Zombies, based on clicks, target ROAS and cost per conversion (Clarmix Product Labelizer). The agent can list the products labeled Costly or Zero Conversion, show what they spent, and check in Merchant Center whether a feed issue explains the result before you exclude anything. If labels are new to you, our explainer on what a labelizer does with custom labels covers the basics. Our post on common Performance Max mistakes covers what to do with them next.

"Help me run a keyword search to improve the performance of my PMax"

Performance Max doesn't use keywords, but it does take search themes: words and phrases your customers use, added on top of the queries Google already predicts. You can add up to 50 per asset group, and exclusions and negative keywords still apply (Google Ads Help: search themes). The agent can start from your best-selling products and their titles, suggest themes per asset group, and flag themes that would overlap.

More questions worth asking

• "Which products got the most clicks last month without a single conversion?"
• "Which Merchant Center issues appeared this week, and how much did the affected products spend?"
• "How much of each Performance Max campaign's spend went outside Shopping?"
• "Summarize this client's last 30 days in five bullet points I can send them."

Each one mixes sources: Google Ads for spend and conversions, Merchant Center for product status, Clarmix for labels and PMax insights. That mix is what turns a general assistant into an e-commerce agent.

How does the Clarmix MCP server work?

The Clarmix MCP server works in Claude and ChatGPT, and in other AI tools that accept MCP. As of October 2026, it is not yet listed in Claude's or ChatGPT's connector directory, so you add it as a custom connector with its server URL. It uses the Google Ads, Merchant Center and Clarmix APIs, so one connection gives the assistant all three sources, with up to 90 days of data. It is read-only: it cannot change anything in your accounts.

It is dedicated to Google Shopping. At Clarmix we trained it for e-commerce, on real use cases from real e-commerce brands, so it knows to look at products, feeds and listing groups rather than only at campaigns. It is built for PPC consultants who manage Google Ads for online stores and want their own agent to help with decisions.

The Clarmix data adds context Google Ads doesn't have on its own: product labels, feed audit scores and PMax insights such as spend outside Shopping.

How do you set up a Google Shopping agent in Claude or ChatGPT?

1. Create a Clarmix account and connect Google Ads and Merchant Center to it. The connection goes through Google's official APIs, and the agent reads only the accounts you link there. The MCP server requires a paid Clarmix plan.
2. In Claude, add Clarmix as a custom connector. In claude.ai or Claude Desktop, open Settings, then Connectors. Clarmix isn't listed in the directory yet, so click "Add custom connector" and paste the server URL: https://api.clarmix.com/mcp
3. In ChatGPT, add the same server URL in Developer mode. Clarmix isn't listed in ChatGPT's directory yet, so go to Settings, then Apps, then Developer mode, and add https://api.clarmix.com/mcp. In another AI tool that accepts MCP, add the same URL as a custom connector.
4. Sign in to Clarmix and approve the connection. The approval screen lists what the assistant will be able to read: Google Ads performance, campaigns and change history; Merchant Center products, issues and feeds; Clarmix labels, rules and enrichment. It cannot change anything in your accounts. You can give access to all your accounts, including future ones, or only to the accounts you pick.
5. Ask in plain language, and check the first answers on an account you know well. Ask about a client account, a campaign or a single SKU, the way you would ask a colleague. The assistant retrieves the Google Ads, Merchant Center and Clarmix data it needs and answers with your numbers.
6. Save the questions that work as a routine: a weekly account review, a costly-products check, a search themes refresh.

What should you check before acting on an agent's answer?

• The date range. Ask the agent which period it used. The Clarmix MCP server retrieves up to 90 days, so year-over-year comparisons need another source.
• The numbers behind the advice. Spot-check a few figures in Google Ads before you change a budget or exclude products.
• Who makes the change. The Clarmix MCP server can't change your accounts, so a person applies every recommendation. Take extra care with exclusions and target changes.
• Conversion delay. Recent days are often incomplete, so treat the last few days of any trend with care.

Frequently asked questions

How do I connect Claude to Google Ads?

Through a connector built on MCP. With Clarmix, you connect Google Ads and Merchant Center to a paid Clarmix account. Then, in Claude, open Settings, then Connectors, click "Add custom connector", paste https://api.clarmix.com/mcp, sign in to Clarmix and approve the connection.

Does it work in ChatGPT too?

Yes. Link your accounts in Clarmix, then in ChatGPT go to Settings, then Apps, then Developer mode, add https://api.clarmix.com/mcp, sign in to Clarmix and approve the connection. It also works in other AI tools that accept MCP.

How far back can the agent look?

Up to 90 days of data, which covers weekly reviews, monthly client reports and a 3-month trend.

Is it the same as Google's Ask Advisor?

No. Ask Advisor is Google's agent built with Gemini and works inside Google Ads. The Clarmix MCP server works inside Claude or ChatGPT and combines Google Ads, Merchant Center and Clarmix data.

Is Google's official Google Ads MCP server enough?

For read-only Google Ads reporting, it can be. It is strictly read-only, covers the Google Ads API only and needs a Google Cloud project. Product questions that also need Merchant Center data need another source.

Sources

• Model Context Protocol: What is MCP?
• Claude Help Center: Use connectors to extend Claude's capabilities
• OpenAI Help Center: Connected apps in ChatGPT
• Google Ads: Ask Advisor
• Google for Developers: Google Ads MCP server
• Google for Developers: Merchant API MCP Access Service (Alpha)
• Google Ads Help: Use search themes with your Performance Max campaign
• Clarmix: MCP server

Cover photo by Luke Chesser on Unsplash.