Google Ads AI Max: How It Works and Best Practices

By Corey Frankosky · Last Updated July 23, 2026

Google Ads AI Max is a setting inside your existing Search campaigns that lets Google match your ads to searches beyond your keyword list, write ad copy from your website content, and send traffic to the pages it thinks convert best. It is not a new campaign type and you do not rebuild anything to use it. The honest summary after using it across client accounts is that it works when you feed it good conversion data and it wastes money when you do not. This guide covers what AI Max actually does, what Google claims versus what independent testing found, the February 2027 Dynamic Search Ads deadline, and the specific settings I would lock down before turning it on.

1What AI Max Actually Is

The simplest description I can give: AI Max is Dynamic Search Ads combined with broad match keywords, under a new name. That is not a criticism, it is just what it does.

I walk through the settings screen by screen in the video above, which is episode 27 of the Surfside PPC podcast. The rest of this post covers the same ground in writing, plus the performance data and deadlines that landed after I recorded it.

It is a setting, not a campaign type. Open a Search campaign, scroll through the settings, and you will find it. Two halves sit underneath it:

• Search term matching. Google matches your ads to queries beyond your keyword list, using your landing pages and assets to work out what you sell. A campaign built around "emergency tree removal near me" starts matching conversational searches like "large oak leaning on my roof after storm."
• Asset optimization. Google writes headlines and descriptions from your website content, and Final URL Expansion sends traffic to whichever page it believes will convert best.

Two changes since it launched that most articles have not caught up with. AI Max is out of beta and fully rolled out, so it is in your account today. And as of mid-2026 it is now the default setting on newly created Search campaigns, with the ability to opt out. It is still opt-in for campaigns you already run. If you build a new Search campaign and do not check that setting, you are probably running AI Max without deciding to.

Here is my slightly cynical read on why Google wants this. AI Max leans heavily on your website content. It reads your pages, finds the ones most likely to convert, and builds ads from them. If you have spent years building a content library, AI Max can go find the searches that library ranks for and charge you for those clicks. In a way you are paying for traffic you might otherwise have earned through SEO. That is the direction Google is heading, keyword targeting keeps getting more limited, and it is worth understanding rather than ignoring.

Question to Answer:

Have you created a new Search campaign in the last few months, and did you check whether AI Max was on by default?

2What Google Claims and What Testing Found

This section matters more than any other, because the gap between the two is where budgets get lost.

Google's claim. Advertisers who turn on AI Max typically see 14% more conversions or conversion value at a similar CPA or ROAS. For campaigns still mostly using exact and phrase match keywords, that figure rises to 27%.

Read the qualifiers, because they are doing a lot of work. The 27% figure comes from Google internal 2025 data, applies to campaigns where more than 70% of conversions came from exact or phrase match keywords, and excludes retail advertisers. That is a narrow slice. If your account already runs broad match with Smart Bidding, you are not the campaign in that study.

What independent testing found. Smarter Ecommerce analysed more than 250 Search campaigns running AI Max and reported a median revenue lift of about 13%, which lines up with Google's number. The part Google does not lead with is what came alongside it: a median CPA increase of roughly 16%, and a ROAS range stretching from 42% above baseline to 35% below. Only around 22% of those campaigns landed anywhere near their original ROAS targets.

Both things are true at once. AI Max can produce more conversions and it frequently produces them at a higher cost. Whether that trade is worth taking depends entirely on whether your conversion data is good enough for Google to find the right people, which is the theme of this whole post.

Question to Answer:

If AI Max delivered 13% more revenue at 16% higher cost per acquisition, would that be a win or a loss for your business?

3What You Still Control

Automation does not mean handing over the account. You keep the inputs that shape everything the system does.

Final URL Expansion deserves particular attention. Turned on, Google can send a visitor to whichever page it thinks is more relevant. Sometimes that is smart. Other times it sends paid traffic to a privacy policy, a careers page, or a thank-you page. Turned off, traffic stays on the URLs you chose.

Question to Answer:

Do you know which pages on your site would receive traffic if Google were choosing the landing page instead of you?

4What I Have Seen in Client Accounts

Here is the honest report from my own accounts, because the results have been mixed and the pattern is consistent.

I turned it on for a dog trainer with a modest budget, mostly as a test. It pulled in completely irrelevant traffic. Searches about cleaning, dog grooming, all sorts of things nowhere near the intent we wanted. Somebody who grooms their dog might eventually need training, but that is not the same as buying intent, and on a small budget you feel that waste immediately.

On other client accounts where we were feeding in good conversion data, it worked noticeably better. That is the whole pattern in one line: good conversion data in, decent results out. Poor or no conversion data in, wasted budget out.

The type of business matters just as much. AI Max is built for companies with volume and clean data flowing back into Google. If you sell makeup or fashion online, where people search in a thousand ways you could never cover with a keyword list, it makes complete sense. If you are a roofer, it is harder. Somebody calls, you send an estimator, they collect three other bids, and the job might close a month later. That long, low-volume journey is difficult to feed back to Google as usable conversion data.

My own business falls into the harder category. I am not landing 30 new clients a month with a steady stream of leads and closings coming through Google Ads. When you do have that, it is much easier to send the data back and say here is our CRM, here are 40 new clients this month, here is where they came from. With that, Google can actually understand what your ads produce.

Question to Answer:

Does your business produce enough conversion volume each month for Google to learn a pattern from, or are you asking it to learn from a handful of leads?

5The Settings I Would Lock Down First

If you do turn it on, these five controls make the difference between a test and a leak.

1. Set brand controls to unbranded searches only. Under brand settings you choose how to handle searches containing brand names. For most accounts, show ads only on unbranded searches. Turning AI Max loose on branded terms is generally a bad idea, and if you already run a brand campaign it will compete with it. Our post on using branded keywords in Google Ads covers why brand belongs in its own campaign.
2. Add URL exclusions. Decide which pages should never receive a paid click. Careers, financing, privacy policy, thank-you pages. Exclude them before launch, not after you see the spend.
3. Turn off Final URL Expansion at launch. Keep traffic on the pages you chose until you have enough data to judge the campaign. Turn it on later if you want to test it.
4. Use locations of interest. A local business can point AI Max at the specific counties and services it actually covers. While you are in there, set location targeting to presence only rather than the presence or interest default.
5. Use text guidelines and disclaimers. Give the AI a brief so generated copy follows your brand rules, and lock required wording in place with text disclaimers where the exact phrasing matters.

Leave Target CPA and Target ROAS blank at launch. Setting a target before the campaign has data can choke delivery down to a fraction of the budget. Add one once you have 15 to 30 conversions to base it on. Our guide to Google Ads bidding strategies covers when each strategy is appropriate.

The goal with all of it is the same. Keep your brand rules intact while letting Google go look for conversions you are not already capturing.

Question to Answer:

Which pages on your site would embarrass you if a paid click landed on them, and have you excluded those URLs?

6Conversion Data Is the Whole Game

Feeding Google quality conversion data is the single most important thing you can do here. Accounts that do not provide good conversion data are the ones that get left behind.

The mistake I see most often is marking soft actions as primary conversions. Page views, PDF downloads, newsletter signups. AI Max optimizes toward whatever you mark primary, so if you tell it a page view is a win, it will go find you page views.

• Set primary conversions to outcomes only. Qualified form submissions, calls over a minimum duration, purchases with revenue values attached.
• Import offline conversions for longer sales cycles. Connect your CRM through Google Click ID so Google learns which leads actually closed rather than stopping at the form fill.
• Attach revenue values for ecommerce. Without values, the system optimizes for conversion count and will happily find you cheap low-value orders.
• Keep secondary actions secondary. Track click-to-call and multi-step form starts so you can see them, without letting them drive bidding.

There is also a volume floor. Forty conversions spread across eight campaigns gives each one a thin signal. The same 40 in two campaigns is a much stronger one. Aim for 30 or more conversions per campaign per month before expecting stable performance, and consolidate rather than fragment. Our post on setting up Google Ads conversion tracking covers the build.

Question to Answer:

How many conversion actions are marked primary in your account, and would you call every one of them a customer?

7The Reporting Problem Nobody Mentions

This one is worth knowing before you measure a test, because it can make AI Max look better than it is.

Testing by Adalysis surfaced a reporting quirk. When AI Max is enabled, search terms that would have matched your existing exact match keywords can get attributed to AI Max in reporting instead. Not because AI Max found anything new, but because it takes credit for matches your keywords would have caught anyway.

The practical effect is that the AI Max line in your report can look productive while the incremental gain is smaller than it appears. If you are running a test to decide whether to keep it on, that matters a great deal.

Two ways to protect yourself. Run AI Max as a campaign experiment with a proper split rather than flipping it on and comparing to last month, since seasonality and everything else in the account moved too. And judge the test on total campaign conversions and total cost, not on the AI Max rows in the search terms report.

Speaking of which, filter your search terms report by AI Max match type every week. This is where you catch irrelevant queries before they add up, and AI Max drives considerably more search term volume than Dynamic Search Ads did. Our guide to building negative keyword lists from the search terms report covers the workflow.

Question to Answer:

If you tested AI Max, did you compare total campaign performance, or did you read the AI Max rows in the report and stop there?

8The February 2027 DSA Deadline

Dynamic Search Ads are being retired into AI Max, and the timeline moved in your favour. On June 11, 2026, Google announced the automatic migration is postponed from September 2026 to February 2027.

The September date is the one people are missing. Only the DSA automigration moved. Automatically created assets and the campaign-level broad match setting still transition in September 2026 as originally planned.

Treat the extension as runway rather than a reprieve. Migrating yourself means you keep control of campaign structure, bidding, and URL settings, and you preserve your historical reporting. Waiting until February 2027 means Google decides, and any DSA campaign still running gets converted without further input from you.

Question to Answer:

Do you still have Dynamic Search Ads campaigns running, and have you tested an AI Max equivalent against them yet?

9The August 17 Bidding Change

This is not strictly an AI Max change, but it lands in the middle of everything above and it is close enough to matter right now.

Starting August 17, 2026, Google is changing how Target CPA and Target ROAS behave on campaigns that are limited by budget. Today a budget-limited campaign can quietly beat its target. You set a $10 Target CPA, Smart Bidding settles at $5, and you keep the difference. After August 17, delivery gets steered back toward the number you typed in.

The scope is narrower than a lot of coverage suggests, so be precise about it:

• Affected: campaigns using Target CPA or Target ROAS that are also flagged as limited by budget.
• Not affected: campaigns with unconstrained budgets, and campaigns already delivering at their target.
• Google will not adjust your targets for you. If your target is stale, your cost per acquisition drifts up toward it.

The audit takes an hour. Filter for campaigns marked "Limited by budget" running a target strategy, pull the last 30 days, and compare actual CPA or ROAS against the target. Anywhere actual performance is meaningfully better than the target, either bring the target in line with reality, raise the budget where the economics support it, or switch to Maximize Conversions or Maximize Conversion Value and accept a floating CPA.

Google has published a Bid Target Adjustment Tool to help with this. Our post on Google Ads budget pacing covers how to tell whether a campaign is genuinely budget-limited in the first place.

Question to Answer:

How many of your campaigns are marked limited by budget and beating their target right now?

10Should You Turn It On

My answer right now: yes, test it for a week or two, but only in accounts that are spending real money, already performing well, and have solid conversion data behind them.

If you are running a brand new campaign with no data, absolutely not. You will waste the budget. And if your conversion tracking is poor or missing, fix that first, because AI Max will simply find you more of the wrong thing faster.

One argument in favour that is worth weighing separately from performance. Google has confirmed that exact match keywords are not eligible to serve inside AI Overviews. Only broad match, Performance Max, and AI Max can serve in that placement. If a meaningful share of your queries now return an AI Overview, an exact-match-only account is locked out of those auctions regardless of how well the keywords perform elsewhere. That is a structural reason to run at least one AI Max test, separate from the conversion numbers.

Question to Answer:

Do your highest-value searches now return an AI Overview, and is your account eligible to appear inside it?