AI Max for Search in Healthcare: What Keywordless Matching Does to Procedure Queries

This is for paid search managers and marketing leads at hospitals, clinics and healthtech companies who have been offered AI Max for Search in an existing campaign, or who have seen it switched on by an upgrade prompt. It explains what the feature does according to Google's documentation, why that matters when the queries are about procedures, diagnoses and symptoms, and how to test it without handing control of sensitive queries to an automated layer.

It does not report test results, because we cannot publish another account's numbers as a benchmark. It gives a method you can run on your own account, and it assumes you already know Search campaign structure and match types.

For the demand side of procedure queries, our piece on ranking for high-intent medical tourism queries shows how organic and paid work together on the same searches.

What AI Max actually is

Google's help page, Optimize Google Ads with AI Max for Search campaigns, describes AI Max as a suite of targeting and creative features that works as an optimisation layer on a Search campaign. It is not a separate campaign type. You switch it on inside an existing Search campaign.

Three components matter for healthcare.

• Search term matching. Turning on AI Max enables search term matching, which uses broad match and keywordless technology to find additional relevant queries. It can be switched off in ad group settings.
• Text customization. Formerly called automatically created assets, it generates ad copy tailored to the user's search, drawing on your existing ads, landing page copy and assets. It is set at campaign level and can be turned off separately.
• Final URL expansion. It sends traffic to the most relevant URL on your site when that is likely to perform better. It depends on text customization, and pinned assets are not respected when a different URL is chosen.

Controls exist: URL exclusions at campaign level, URL inclusions at ad group level, brand inclusions and exclusions, and locations of interest. Reporting adds an "AI Max" match type and a source column in the search terms report, a "Selected by" column in landing pages, and combined views of search terms, headlines and URLs.

Why procedure queries are different

In most retail accounts, a wrong match costs a click. In healthcare, a wrong match can also place a promotional message next to a worried person's query about a diagnosis, a fertility problem or an intimate health concern. Keywordless matching widens the set of queries your ad can enter, starting from your landing pages and ads rather than from keywords you chose. That is useful when the hospital has a large service catalogue that nobody has mapped to keywords. It is a risk when the catalogue contains pages about sensitive conditions.

There are four specific exposure points.

1. Query scope. The system may match queries you never wrote a keyword for, including symptom and condition language that your clinical team would not want to answer with an ad.
2. Generated headlines. Text customization writes new headlines and descriptions. Generated wording about a procedure is health content. It needs the same review as any ad copy, and the clinical owner is the person to approve it, which is hard if the text is created at serve time.
3. Destination choice. Final URL expansion may select a different page than the one you chose, including pages for doctors, packages and locations. Our companion teardown on landing URLs covers that in depth.
4. Policy. Google's healthcare and personalised advertising policies still apply to every ad. Health is a sensitive interest category under the personalised advertising policy, which restricts advertiser-curated audiences for health ads. AI Max does not change what you are allowed to say or whom you may target. Read the current policy before you test, and confirm whether your ads fall in a restricted class.

For India, the NMC rules on professional advertising apply to what doctors may say about themselves, and a generated headline that praises a named doctor is the kind of wording that needs a compliance check. The NMC compliance handbook sets out the framework.

Decide whether a campaign is a candidate

Not every healthcare campaign should receive AI Max. A rough triage, to be adjusted with your compliance owner:

The sensitive categories are not off limits forever. They are not the right place to learn how the system behaves.

A safe test design

Google's own guidance is that AI Max is less effective when a campaign is limited by budget, and it flags that. So choose a campaign with headroom. Then build the test so that you can compare, not just observe.

1. Pick a candidate campaign with a stable conversion history and tracked enquiries, not only page views.
2. Create a clean comparison. Use a campaign experiment in Google Ads where available, or a duplicated campaign with a split budget, so the same demand is exposed to both versions. If you split by time instead, seasonality will spoil it.
3. Turn on search term matching first and leave text customization and final URL expansion off for the first phase. That isolates the query expansion effect. Because final URL expansion depends on text customization, switching text customization off also keeps final URL expansion off.
4. Set URL exclusions for pages that should never be a landing page for paid traffic: careers, news, doctor pages that fail compliance review, outdated package pages.
5. Set brand controls to exclude competitor or irrelevant brands, using the brand settings described in Google's brand lists help page.
6. Fix a review date after a full conversion delay window, and record the pre-test baseline for qualified enquiries, not only form fills.

In a second phase, if the first behaves, turn on text customization and review the asset report weekly for generated wording. Add final URL expansion last, with exclusions in place.

What to read in the reports

The search terms report with the AI Max match type shows the queries the new layer introduced. Export them and classify each query manually into four buckets:

• relevant and desirable (a procedure or service you offer);
• relevant but sensitive (needs a copy or landing page decision);
• irrelevant (add as negatives);
• informational (people seeking information with no booking intent).

A hypothetical illustration of the classification only: for a dental group, "dental implant cost near me" would be desirable, "bleeding gums causes" informational, and "dental college admission" irrelevant. The proportion in each bucket is the finding, and it will differ by account. Do not copy another account's split.

Then examine the cost of the added queries against the qualified enquiries they produced. If AI Max added queries that convert at an acceptable cost per qualified enquiry, keep them. If it added mostly informational or irrelevant queries, tighten with negatives or turn the matching off.

Do the same with the landing pages report and the "Selected by" column once final URL expansion is on: confirm that the pages chosen are the ones a patient would want. If the selected pages are poor, exclude them, or fix the pages.

Governance for generated text

Healthcare marketing needs a named reviewer for ad copy. Generated headlines break the usual workflow, in which every line is approved before launch. Two practices help. First, keep a short list of phrases never to use (outcome promises, comparative claims, "best", "cure"), and check the asset report against it each week. Second, remove text customization from any campaign where a reviewer cannot keep that pace. The feature can be turned off separately, so you do not have to give it up entirely to use search term matching.

Tracking templates also deserve a test. Google's page warns that dynamic landing pages can break with certain tracking templates and recommends testing landing pages before enabling AI Max. In a hospital site with location-specific URLs and redirects, a broken template can send paid traffic to error pages.

Where to start

1. Choose one low-sensitivity, budget-healthy campaign.
2. Record the baseline: qualified enquiries, cost per qualified enquiry and top search term categories.
3. Run search term matching alone for one full conversion cycle.
4. Classify the new queries and add negatives.
5. Add text customization with weekly copy review, then final URL expansion with exclusions.
6. Keep sensitive-condition campaigns outside the test until you trust the process.

Our Google Search ads service and Google Ads audit include this staged approach to automated features in healthcare accounts.