AI Max for Google Search Ads in 2026: How to Test Query Expansion, Generation and Landing Page Selection

AI Max has become one of the most important changes to Google Search advertising in 2026 because it extends automation beyond bidding and conventional keyword matching. It can broaden the searches that trigger an advert, create additional headlines and descriptions, and select a different page from the advertiser’s website when that page better matches the user’s intent. These capabilities can increase reach, but they also change several variables at the same time. For marketers, the sensible approach is therefore not to activate every setting and judge the campaign by total conversions a few days later. AI Max should be tested in stages, with a stable baseline and clear success criteria for query quality, advertising costs, generated copy and landing-page relevance.

What AI Max Changes in Google Search Campaigns in 2026

AI Max is not a separate Google Ads campaign type. It works as an optimisation layer for Search campaigns and combines several functions that previously had to be managed more independently. One of the central functions is search term matching. This extends reach beyond the advertiser’s existing keywords by using broad-match and keywordless matching technology together with information from keywords, advertisements and website pages. In practical terms, Google can identify searches that were not explicitly covered by the original keyword list if its systems consider those searches relevant to the campaign. This is why the feature is better viewed as intent expansion rather than simply another version of broad match.

The second important area is asset optimisation. Text customisation, previously called automatically created assets, can produce additional headlines and descriptions from the information already available to Google. Its inputs can include the campaign’s existing advertisements, keywords, domain and landing-page content. Google combines extracted website information with generative AI, which means the resulting text does not always have to reproduce an existing sentence word for word. This can make an advert more closely aligned with an individual search, but it also means marketers need to check whether automatically created statements accurately reflect prices, availability, product characteristics, delivery terms and other information that matters to customers.

Final URL Expansion adds another layer. Instead of sending every click to the Final URL originally entered in the responsive search advert, Google can select another relevant page from the same domain when it predicts that the alternative page better answers the search. Advertisers can use URL exclusions and other controls to prevent unsuitable sections from being selected. AI Max also provides more detailed reporting for these automated decisions. Search-term reports can identify AI Max traffic and show the source of a match, while landing-page and asset reports help marketers see which pages and generated elements were actually used. This additional visibility is essential because campaign totals alone cannot show whether increased conversions came from better queries, different advertising text or different destination pages.

Why Query Expansion Needs a Controlled Test

Query expansion should be treated as a reach test rather than as permission to accept every additional search Google can find. A campaign may gain more impressions and clicks while attracting people whose intentions are weaker than those of the original audience. For example, an advertiser targeting commercial searches for a paid professional service might suddenly receive additional informational queries from users who are still researching the subject. Those searches are not necessarily irrelevant, but their economic value can be very different. The test therefore needs to measure whether incremental traffic produces useful business results rather than simply increasing traffic volume.

A suitable test campaign should already have dependable conversion measurement and enough recent activity to establish a meaningful baseline. Google also advises against judging AI Max in campaigns that are limited by budget, because a restricted budget can prevent the system from using additional opportunities properly. Conversion-based Smart Bidding is also important for the main AI Max functions, as the system needs conversion signals to decide which new searches are worth entering. Before testing, marketers should verify that important actions such as completed purchases, qualified leads or confirmed bookings are being measured correctly. If weak actions such as page views are treated as primary conversions, AI Max may optimise towards activity that looks successful in Google Ads but has little commercial value.

Google now provides dedicated AI Max experiments for Search campaigns. Instead of requiring a complete copy of the campaign, this experiment can divide traffic and budget within the existing campaign. One part acts as the control with AI Max disabled, while the trial portion operates with AI Max enabled. This reduces some of the differences that can arise when two independently managed campaigns are compared. It also makes the result easier to interpret because both sides remain connected to the same underlying campaign. Marketers should still check campaign eligibility before starting, because factors such as an existing active experiment, shared budgets, certain bidding configurations or previously enabled text customisation can affect whether the dedicated experiment can be created.

How to Test AI Max Query Expansion Without Losing Search Intent

The first useful experiment is usually a test of search term matching with as few other changes as possible. When a new AI Max experiment is created, Google can activate search term matching and asset optimisation for the trial. However, advertisers have controls that allow them to limit which AI Max features participate in the test. If the immediate question is whether expanded matching can find profitable searches, keeping the creative and destination-page variables as stable as possible makes the result easier to understand. Otherwise, a stronger trial may be caused by better query coverage, generated copy, a different landing page or a combination of all three, leaving the marketing team with no clear explanation for the change.

The evaluation should start with business metrics rather than click metrics. Conversion rate, cost per acquisition, conversion value and return on advertising spend are generally more informative than impressions or click-through rate when the objective is sales or leads. A rise in impressions is expected when matching expands, so it should not be treated as proof that the experiment is working. The useful question is whether additional searches generate incremental conversions at an acceptable cost without damaging the economics of the campaign as a whole. For businesses that qualify leads offline, importing qualified or completed lead outcomes into Google Ads can make this analysis considerably more meaningful than optimising only towards the initial form submission.

Marketers should also compare the trial with the campaign’s historical search behaviour. If the campaign previously depended heavily on a small group of exact or phrase-match terms, a larger proportion of new searches is normal after AI Max is activated. What matters is the relationship between those searches and the underlying offer. Some unexpected queries may reveal valuable customer language that was missing from the account, while others may indicate that the system has moved too far from the intended market. This distinction is especially important for businesses with several products, service levels or customer groups, where a term that appears relevant at first glance may lead to an unsuitable offer or a page designed for a different audience.

How to Judge Search Terms and Match Quality

The Search terms report should become part of the regular AI Max review process. Google identifies incremental traffic associated with AI Max and provides additional information about how those searches were matched. The Source column can help distinguish expansion related to broad match from keywordless matching. A dedicated AI Max view can also connect a search term with the headline and landing page shown to the user. This is more useful than reviewing the query by itself. A search that appears slightly broader than expected may still be valuable if the advert and destination page provide a precise answer, while a seemingly relevant search can perform poorly if the accompanying message creates the wrong expectation.

Query quality should be judged by patterns rather than isolated examples. One unusual search with no conversion is rarely a reason to restrict the campaign immediately. Repeated clusters of searches with the same weak intent deserve more attention. Negative keywords remain useful where a clearly unwanted meaning appears consistently, while AI Max also provides controls related to brands and geographical intent. These controls can be particularly important when a campaign must avoid certain brand associations or when the business only serves people interested in specific locations. The purpose is not to rebuild an enormous manual keyword structure around automation, but to set boundaries where the business has information that Google cannot infer reliably from performance data alone.

The test should also be allowed to accumulate enough evidence before a decision is made. Daily changes to exclusions, bidding, budgets and campaign settings make a clean comparison difficult because the conditions of the experiment keep changing. A better approach is to define the primary metric before launch, watch for serious relevance or compliance problems during the test, and postpone routine optimisation until a useful amount of conversion data has accumulated. The appropriate duration depends on traffic and conversion volume rather than a universal number of days. A high-volume retailer can obtain evidence much faster than a business generating a handful of valuable enquiries each week, so statistical confidence and business significance matter more than an arbitrary testing calendar.

How to Test Generated Ad Copy and Final URL Expansion

Once search term matching has been assessed, the next stage is to examine text customisation. Google can create additional responsive search ad headlines and descriptions using the existing campaign and website as source material. Before switching it on, the advertiser should therefore review the pages that Google is likely to read. Old prices, discontinued products, vague claims and inconsistent terminology can become more serious when they are used as inputs for automated advertising copy. Website content should accurately describe the offer before it is used to support generation. After activation, Google-created assets can be identified in reporting, allowing the marketing team to compare their performance and wording with advertiser-supplied assets.

Performance is only one criterion when generated copy is reviewed. Advertisers should also check factual accuracy, brand terminology and the relationship between the promise in the advert and the information available after the click. A generated headline may achieve a strong click-through rate because it is very specific, but that is not useful if the corresponding service has conditions that are missing from the landing page or if the wording suggests something the company cannot guarantee. Google currently offers text guidelines for AI Max Search campaigns as an experimental beta. Where the feature is available, marketers can use term exclusions and messaging restrictions to place firmer boundaries around automatically created text. These controls are useful for requirements that can be written as clear rules, although human review remains necessary.

Final URL Expansion should then be assessed as a separate business decision rather than merely another switch. It requires text customisation because automatically selected destination pages may need different advertising copy from the original Final URL. The potential advantage is straightforward: a broad product or service campaign can send a user directly to the page that most closely matches a detailed search instead of forcing everyone through the same generic page. The risk is equally clear. Pages intended for recruitment, investor information, old campaigns, support documentation or unrelated audiences may be unsuitable destinations even when they contain related words. URL exclusions should therefore be reviewed before launch, and the landing-pages report should be checked afterwards to confirm where AI Max is actually sending visitors.

A Practical AI Max Testing Framework for Marketers in 2026

A manageable testing sequence starts with measurement rather than automation. The marketing team first records the current campaign’s conversion volume, CPA or ROAS, conversion rate, search-term quality and most important landing pages. It then runs an AI Max experiment focused primarily on search term matching and compares incremental queries with the existing traffic. If those searches produce acceptable business results, asset optimisation can be assessed with the same emphasis on controlled change. This staged method is slower than activating every feature simultaneously, but it produces information that can be reused across other campaigns because the advertiser learns which part of AI Max actually created the improvement.

The next stage combines generated text with landing-page selection and checks the complete path from query to advert to page. Reporting should answer three separate questions: did AI Max enter searches that the previous setup missed, did its generated message accurately represent the offer, and did it select a destination page that helped the visitor complete the intended action? A favourable campaign-level CPA can hide problems in one of these areas. For example, strong branded traffic may compensate for weak expanded searches, or a few high-value conversions may make unsuitable landing-page choices less visible in the overall result. Reviewing the underlying reports prevents aggregate performance from masking those issues.

This testing discipline will become more relevant as Google consolidates older Search automation into AI Max. As of August 2026, Google says campaigns using automatically created assets and the campaign-level broad match setting are scheduled to begin automatic upgrades to AI Max from September 2026. The Dynamic Search Ads transition follows a different timetable: Google extended the DSA sunset and now says automatic upgrades are due to begin in February 2027. Marketers therefore still have time to understand how AI Max behaves before some legacy settings are moved into the newer framework. The strongest preparation is not simply enabling automation early, but building reliable conversion data, cleaning website content, defining business boundaries and establishing a repeatable experiment process that shows whether additional reach produces additional value.