What to Let Google's AI Run in Your Ads Account, and What to Keep Your Hands On

Google spent the first half of 2026 putting an AI agent inside every advertiser’s account. Ask Advisor launched in beta at Google Marketing Live in May. AI Max for Search is now a toggle on campaigns most people never asked for. If you run a business and pay for Google Ads, someone is going to tell you this changes everything. Some of it does. Most of it is a decision about who is accountable for your money.

Here is how we sort it.

What Ask Advisor actually is

Ask Advisor is a conversational agent built on Gemini that lives in the top-right corner of your Google Ads account. You ask it questions in plain English and it reads your data to answer. What changed week over week. Why a campaign lost conversions. Why an ad got disapproved and what to change. It will draft new headlines and images, suggest keywords, and walk you through billing and verification problems. Google is connecting it to Analytics and Merchant Center so it can pull from all three without you switching tabs.

Two things matter about how it works. First, it recommends, you approve. Google’s own help page puts the responsibility on you: it is your job to make sure a suggestion is accurate and relevant before you implement it. Second, it is still in beta, English-language accounts only, and it does not work inside manager accounts, so anyone running multiple clients is using it one account at a time.

Our take: Ask Advisor is a very good analyst and a very average strategist. It is fast at the questions that used to eat an afternoon. Pull the week, explain the drop, find the policy problem, draft five more headlines. We use it for exactly that. We do not let it decide where the next dollar goes, because it does not know your margins, your capacity, or that your best month is always March. It knows your account. That is not the same as knowing your business.

What AI Max actually does

AI Max for Search is different. It is not an assistant. It is a set of automations that change how your search campaigns match, write, and land.

Search term matching expands your keywords with broad match and keywordless matching, so your ads show on queries you never bid on. Text customization rewrites your ad copy on the fly using your existing ads, your landing page, and generative AI. Final URL expansion sends people to whichever page on your site Google thinks will convert best, not necessarily the one you chose. All three switch on by default when you enable AI Max. Each can be turned off on its own.

Google gives you a few controls that matter. Brand inclusions and exclusions, so the machine doesn’t spend your non-brand budget on people already searching for you. Locations of interest, so keywordless matches still respect geography. URL exclusions, so it can’t land traffic on your careers page. And the search terms report now labels AI Max matches with their source, so you can see what it found.

On results, be careful with the headline numbers. Google’s own claim has drifted from 14 percent more conversions to 7 percent. Independent tests published this year run the full range. One head-to-head put AI Max cost per acquisition at more than double phrase match. Others report meaningfully lower return on ad spend against traditional match types. The honest summary is that the same settings that lift one account inflate cost per lead on another, and nobody can tell you which yours is until you test it.

Where we draw the line

For most of the businesses we work with, which are lead generation accounts doing somewhere between 20 and 200 leads a month, this is the split.

We let AI do the analysis, the drafting, and the diagnosis. Ask Advisor for the weekly read, the disapproval fix, the asset variations. That work used to belong to a junior buyer. It is faster now and just as good, as long as a senior person reads the output before anything changes.

We keep our hands on anything that spends money in ways we can’t see coming. That means search term expansion stays off in low-volume accounts. A local electrician doing 40 leads a month cannot absorb a learning phase that doubles cost per lead for three weeks. Final URL expansion stays off unless every page on the site was built for paid traffic, which is almost never true. Brand exclusions stay on whenever a separate brand campaign exists. And negatives get reviewed weekly, because expansion without hygiene is just leakage with a nicer name.

How to test AI Max without lighting money on fire

Pick one campaign, not the account. Turn on AI Max but switch off final URL expansion before it runs. Add your brand terms as exclusions. Set locations of interest to your actual service area. Leave it four to six weeks so the learning phase clears. Compare cost per lead against the same campaign’s prior period and against a campaign you didn’t touch. Read the search terms report every week and add negatives as you go. If cost per lead is still up more than 20 percent after learning, turn it off and move on. You lost a little money and learned something. That is a cheap education.

The point underneath all of it

Google’s AI is optimizing for Google’s objective function inside the boundaries you set. Where the boundaries are loose, it will spend. That is not a criticism. It is a description of the product. The job of whoever runs your account is to set the boundaries, read what the machine did, and decide what happens next. AI made that person faster. It did not make them optional.

If you want a second opinion on what’s switched on in your account right now, that’s a short conversation. Get in touch.