AI Agent for Google Ads: What Ask Advisor and MCP Can Do

Choosing an AI agent for Google Ads used to mean handing your login to an outside tool and hoping for the best. That has changed. Google now builds an agent into the Google Ads interface, and it also offers an official, open-source way for other AI assistants to read your data.

So this guide is for small-business owners, freelancers and in-house marketers who run their own campaigns. It explains what each option can do as of October 2026, how to set it up and what it costs. Along the way, it covers the main risks. It is based on publicly documented capabilities, and TheJusGrow has not tested these products for this article.

What can an AI agent for Google Ads actually do?

There are three realistic routes today, but they suit different people. Each one also gives the AI a different level of control over your account.

Route	What it is	Can it change your account?	Best for
Ask Advisor (beta)	Google’s chat agent inside Google Ads	Yes, with your approval, for some edits	Most advertisers
AI Max and other automation	Google’s AI features built into campaigns	Yes, within the settings you choose	Advertisers who want reach and creative help
Google Ads MCP server	An official bridge for MCP-compatible assistants	No, it is read-only	Analysts and technical users

So “AI agent” can mean a chat assistant, a campaign feature or a data connection. Therefore, the useful question is not which one is smartest. Instead, ask what you want the AI to do, and how much you want it to touch.

Ask Advisor: Google’s built-in AI agent for Google Ads

Google announced Ask Advisor on May 20, 2026, at Google Marketing Live. According to Google’s announcement, it orchestrates a team of expert agents across Google Ads, Google Analytics and Google Marketing Platform. Google also said a Merchant Center agent would join later.

Meanwhile, Google’s help page describes Ask Advisor as an agentic conversational experience built with Gemini. To open it, first select the Ask Advisor icon in the top-right of Google Ads, which carries a Beta badge. Alternatively, you can type a question into the Google Ads search bar. A chat panel then opens on the right.

According to the help page, Ask Advisor can help with:

• Reporting: week-over-week performance, conversions by device and charts, such as impressions over 30 days.
• Troubleshooting: performance drops, policy disapprovals, appeals and final URL updates.
• Account admin: identity verification, billing, failed payments and promo code issues.
• Optimisation: Performance Max suggestions, asset audits, keyword ideas and new campaign ideas.
• Creative: new text headlines and images, which come from your landing page or from Google AI.

It can also save business preferences under “saved info”, so you do not have to repeat context each time. In August 2026, Google also added AI insight cards and prompt-built dashboards to Google Ads, again in beta for English-language accounts. Before Ask Advisor, Google ran separate Ads Advisor and Analytics Advisor tools. However, industry site PPC Land reported that Ask Advisor replaces them with one agent.

What can Ask Advisor change in your account?

This, in fact, is where an assistant becomes an agent. Google’s help page lists campaign edits it can make: pausing campaigns, renaming them, changing targeted locations and updating daily budgets. Crucially, the page says you can implement suggested changes “with your approval”.

That approval step is your main safeguard when you use any AI agent for Google Ads, so treat it seriously. Read each proposed change before you accept it. For example, check which campaign it affects, the old and new values, and whether the change applies today or later.

Google also says you keep full editorial control over AI-generated assets and should always review suggested content. Similarly, the help page tells you to verify billing status, identity verification documents and policy appeal outcomes in your account settings, rather than relying on the chat.

There are limits, too. Ask Advisor is in beta, so Google warns there may be limitations. For instance, it is not available in manager (MCC) accounts, which matters for agencies. Google’s blog also says it is available in beta for English-language accounts. So far, Google has not given a timeline for other languages.

AI Max and other automation: when the campaign is the agent

However, not every AI feature in Google Ads is a chat window. Instead, some of the most powerful automation sits inside the campaign settings themselves.

AI Max for Search campaigns is a good example. Google describes it as a suite of targeting and creative enhancements. It expands your keywords to reach more searches and customises headlines and descriptions for each ad. It can also choose landing pages that match what people want.

Google’s help page cites internal 2025 data on the results. It says advertisers who use AI Max typically see 14% more conversions or conversion value at a similar cost per acquisition or return on ad spend. That is Google’s own figure, however, and your results will depend on your business, budget and tracking.

Google is also testing Direct Offers, a beta promotion format for AI Mode. According to Search Engine Land, these offers only serve on campaigns that use AI Max or text customization.

The trade-off is control. When Google’s AI chooses search terms and writes ad text, you see less of each decision in advance. So it pays to review search term reports and generated assets regularly, especially in the first few weeks.

Google’s AI features also affect where ads appear. Our guide to ads in AI search explains how campaigns can show inside Google’s AI Overviews and AI Mode.

Using the official Google Ads MCP server with an AI assistant

Finally, the third route is for people who want to analyse their account with a general AI assistant or coding tool. Google publishes an open-source Google Ads MCP server that connects the Google Ads API to any assistant that supports the Model Context Protocol.

The most important detail is that the server is strictly read-only. In fact, Google’s documentation says it cannot modify bids, pause campaigns or create assets. As a result, it is a safe way to ask questions, but not a way to automate changes.

The server exposes three tools. First, one lists the accounts you can access. Second, another runs searches written in Google Ads Query Language, which return metrics, budgets and status. Third, a final tool returns metadata about a resource type, such as which fields a campaign has.

In plain English, you can ask an assistant which campaigns spent the most last month. It then writes and runs the query for you. If MCP is new to you, our explainer on the Model Context Protocol covers how these connections work.

How to set up the read-only MCP server

Setup is technical, so budget an hour if you have not used Google Cloud before. Still, it is the most flexible way to build a read-only AI agent for Google Ads analysis. Based on Google’s documentation, last updated on September 30, 2026, the steps are:

1. Create or choose a Google Cloud project that has Google Ads API access at the Explorer, Basic or Standard level.
2. Set up OAuth credentials, either an OAuth client ID and secret or application default credentials.
3. Add a google-ads-mcp entry to your AI tool’s MCP settings. For example, Google’s sample runs the Python server with pipx and sets your project ID and credentials file.
4. If needed, add a login customer ID if you need access through a manager account.
5. Restart your assistant, then test it with a simple request, such as listing your accessible accounts.

Google’s page also notes that developer tokens are no longer required, because they were sunset on September 9, 2026. Instead, your Cloud project’s access level is what counts. For teams, Google also documents an optional Cloud Run deployment, although you must rebuild the image to pick up updates.

Five workflows to try with an AI agent for Google Ads

First, start with low-risk jobs where you can easily check the answer. Here are five that suit an AI agent for Google Ads, using the documented features above:

1. Monday check-in: ask Ask Advisor to compare last week with the week before and explain the biggest change.
2. Disapproval fix: ask why an ad was disapproved and what policy applies, then make the fix yourself.
3. Wasted spend review: through the MCP server, ask an assistant to list search terms with clicks but no conversions.
4. Fresh headlines: ask for new headline ideas, then edit them for accuracy and brand voice.
5. Budget proposal: ask Ask Advisor to suggest a daily budget change, and approve it only after checking the numbers.

Also, a clear instruction gets a better result. Here is an example you can adapt:

Compare the last 14 days with the previous 14 days for my Search campaigns. List the three campaigns with the biggest change in cost per conversion, explain the likely reasons, and suggest one change for each. Do not apply any changes.

Notice the last sentence. After all, saying what the agent must not do is just as important as saying what it should do. Our guide on how to write instructions for AI agents has more examples.

Risks: spend, permissions and prompt injection

Google Ads spends real money, so an AI agent for Google Ads can make costlier mistakes than most apps. Keep these risks in mind:

• Unchecked budget changes: an approved edit can raise spend immediately. Therefore, confirm the amount and the campaign every time.
• Too much access: give outside tools the narrowest access that works. For that reason, the read-only MCP server is a sensible default for analysis.
• Prompt injection: if an assistant also reads web pages or files, hidden text could try to steer it. Even so, limited access and human review remain the best defence.
• Confident errors: AI explanations can sound certain and still be wrong. So check figures against your own reports before acting.

For background on that third risk, see our explainer on what prompt injection is. In addition, review which third-party apps can reach your Google account from time to time, and remove any you no longer use.

Who should use an AI agent for Google Ads, and who should skip it?

Good fit	Probably skip for now
Owners who manage their own small account	Agencies that work mainly in manager (MCC) accounts, where Ask Advisor is not available
Marketers who want faster reports and explanations	Teams that need full automation of changes through outside tools
Analysts comfortable with Google Cloud setup	Anyone without conversion tracking set up, since the advice depends on it
Advertisers in English-language accounts	Accounts in other languages, until Google expands the beta

Costs are simple to describe. Google’s help page does not list a separate fee for Ask Advisor, and the MCP server is open source; you still pay for your ads as usual. However, if you host the server on Cloud Run, check Google Cloud pricing first. For a wider view of how agent plans and credits work, see our guide to AI agent pricing.

A short checklist before you start

• Confirm that conversion tracking works, because every recommendation depends on it.
• Decide which changes you will let the AI propose, and which you will always make yourself.
• Then open Ask Advisor and ask read-only questions for a week.
• Approve edits one at a time, and note what changed and why.
• Use the read-only MCP server for deeper analysis, rather than granting write access to outside tools.
• Finally, review search terms, generated assets and spend weekly.

AI agent for Google Ads: FAQs

Sources

• Google Ads Help, “Ask Advisor in Google Ads (beta)” (accessed October 2026)
• PPC Land, report on Ask Advisor unifying Google Ads, Analytics and commerce (May 2026)
• Google Ads & Commerce Blog, Ask Advisor announcement (May 20, 2026)
• Google Ads & Commerce Blog, AI and agentic updates for Google Ads and Google Analytics (August 10, 2026)
• Search Engine Land, “Google Ads tests new Direct Offer asset for AI Mode promotions” (October 7, 2026)
• Google for Developers, Google Ads API MCP server documentation (last updated September 30, 2026)
• Google Ads Help, “About AI Max for Search campaigns” (accessed October 2026)