Google Ads AI Max can expand a Search campaign beyond its existing keywords, create new text assets, and send traffic to additional landing pages.
That can open up useful demand. It can also change several parts of a working campaign at the same time.
I would not treat that as a reason to ignore AI Max. I would treat it as a reason to test it properly.
Google now offers a built-in AI Max experiment that compares a control with AI Max turned off against a trial with AI Max turned on inside the existing campaign. That gives advertisers a cleaner way to answer the practical question: did the extra automation produce better business outcomes, or did it simply find more traffic?
AI Max changes more than keyword matching
AI Max is not a separate campaign type. It is an optimization layer for an existing Search campaign.
Its two main feature groups are search term matching and asset optimization.
Search term matching uses broad match and keywordless technology to reach queries beyond the campaign’s existing keywords. Google says the system learns from the keywords, creative assets, and URLs already in the campaign.
Asset optimization can include text customization and Final URL expansion. Text customization may create new headlines and descriptions. Final URL expansion may select a different page on the advertiser’s site when Google expects it to be more relevant to the search.
That means an AI Max test is not only a test of broader targeting. It may also be a test of what the ad says and where the click lands. Google’s AI Max documentation explains the available features and controls.
The experiment is the useful part
The most practical AI Max development is not the promise of more reach. It is the ability to test the feature without rebuilding the campaign as a separate copy.
Google says the experiment diverts a percentage of the existing campaign’s traffic and budget into two groups:
- Control: AI Max remains off
- Trial: AI Max is turned on
Both groups remain inside the same campaign. Google positions this as a way to reduce setup differences, shorten the learning period, and get results more quickly than a traditional copied campaign experiment. Those are Google’s claims, not guarantees, but the structure is useful.
In August 2026, Google also announced that AI Max experiments could include brand and location controls. It began rolling out multi-campaign tests for different budgets and ROI targets in September. Availability can vary by account, so I would confirm which options are actually present before planning the test. Google’s August 20, 2026 update
Start with a campaign you can judge
A campaign is not ready for an AI Max experiment just because the button is available.
I would start with a Search campaign that has reliable conversion tracking, enough activity to support a comparison, and a clear definition of success.
That last point matters. If the campaign records every form fill as a conversion but no one is reviewing lead quality, the experiment may optimize toward a bigger number without proving that the leads are useful.
Before starting, I would check:
- Which conversion actions are included in bidding
- Whether those actions represent the outcome the business actually wants
- Whether offline or downstream outcomes can be connected back to the campaign
- Whether the campaign has enough recent volume for a meaningful comparison
- Whether another major change is already affecting the campaign
If tracking is still being repaired or the offer is changing next week, I would fix that first. A clean experiment needs a stable question.
Choose the automation you are actually testing
AI Max can bundle several changes together, but the experiment setup allows advertisers to configure granular asset-optimization controls.
I would make a deliberate decision about text customization and Final URL expansion instead of accepting the defaults without review.
Text customization
This lets Google create new text assets based on the landing page, ads, and keywords.
It may improve relevance for queries the existing ad was not written to address. It can also produce copy that is technically relevant but not how the business wants to describe the product.
Review the pages Google will use as source material. Make sure the important product language, proof points, and claims are accurate before asking the system to create new variations from them.
Final URL expansion
Final URL expansion is enabled by default when an advertiser opts into AI Max, according to Google. It can send a click to a page other than the ad’s original final URL.
That may be useful on a site with clear, focused product pages. It can be risky on a site with old pages, support content, recruiting pages, or loosely related resources.
I would review the site first and add URL exclusions for pages that should never receive paid traffic. Google’s Final URL expansion documentation
Keep the guardrails that matter
More automation does not have to mean removing every control.
AI Max supports brand settings that can specify brands the campaign should be associated with or prevent ads from appearing alongside selected brands. Locations of interest can also be used at the ad-group level to reach searches showing geographic intent.
The right guardrails depend on the campaign.
A non-brand campaign may need brand exclusions so the trial does not manufacture better results by taking credit for branded demand. A location-specific service may need tighter geographic intent controls. A regulated advertiser may need a more conservative approach to automatically generated copy and landing pages.
The point is not to constrain the test until AI Max cannot do anything. It is to prevent the test from winning for the wrong reason.
Review queries, ads, and landing pages together
Do not evaluate the experiment from the headline conversion number alone.
Google provides a Search terms and ad combinations report that connects the query, headline, and landing page used in the customer journey. The keywords report also separates AI Max expanded matches from landing-page-based matches.
I would review:
- Which new search themes are receiving spend
- Which terms are producing qualified conversions
- Which headlines Google is combining with those terms
- Which landing pages are receiving the clicks
- Whether brand traffic is distorting the comparison
- Whether the trial is finding genuinely new demand
Google also allows advertisers to exclude an underperforming search term or landing page directly from the relevant report. Google’s AI Max reporting guide
Measure the outcome, not just the expansion
More queries, clicks, and conversions can look like success. They are not enough on their own.
I would compare the control and trial on:
- Qualified conversion volume
- Cost per qualified conversion
- Conversion rate after the landing page loads
- Lead quality or revenue when that data is available
- Incremental non-brand demand
- Search-term and landing-page relevance
If the trial produces more form fills but the sales team says the leads are worse, that is part of the result. If it finds a few valuable search themes that were missing from the account, that can be useful even if the entire feature suite is not ready to roll out.
The experiment should help decide what to keep, what to exclude, and what to test next.
A practical first test
My starting approach would be:
- Choose one stable Search campaign with reliable conversion tracking.
- Confirm which conversion action represents a meaningful outcome.
- Review the site before enabling Final URL expansion.
- Set brand, location, and URL guardrails that prevent an easy but misleading win.
- Configure the AI Max experiment rather than turning the feature on across the campaign immediately.
- Review search terms, ad combinations, and landing pages during the test.
- Compare qualified outcomes before applying the experiment.
AI Max does not need to be an all-or-nothing decision.
Give the system room to find something new. Keep enough visibility to understand how it found it. Then scale the parts that produce useful business results.