AI Max sounds like the next thing to buy. It's the newest layer Google has added to Search, it arrives already enabled in every campaign you create, and the pitch is simple: more patients from searches your keyword list never imagined. For some accounts, at the right time, it does exactly that.
For a medical practice, the timing is everything. Strip away the name, and AI Max is mostly broad match, plus a machine that writes ad text from your website, plus a feature that picks which of your pages to send people to. Every one of those needs an account that already knows what a patient looks like. In a new healthcare account, AI Max has nothing to learn from and a website full of sentences nobody meant as ad copy.
AI Max Is Three Switches, and the First One Is Broad Match
AI Max isn't one setting. Google lists three features inside it, and each can be turned on or off:
- Search term matching. Google says it uses broad match, along with information from your ads and landing pages, to decide which searches your ads can show for.
- Text customization. Google writes additional headlines and descriptions using content from your website, your landing pages, your existing ads and your keywords.
- Final URL expansion. The landing page stops being yours to pick. Google chooses, search by search, among the pages on your domain.
The first switch is the one that matters most. The method this guide teaches starts every account with Isolation, an exact-match campaign where every search is one the practice chose. Search term matching is broad match by another name. As Chapter 9 warned, a new Search campaign comes with AI Max turned on unless someone turns it off, which quietly turns an exact-match launch into a broad-match one. In healthcare, that's expensive in a specific way. Chapter 6 listed the usual intruders in a medical account: symptom searches from people looking for home remedies, searches for a different specialty that shares a word with yours, job and training searches, other practices' names. Broad match finds all of them, and a new account has no booking history to teach it the difference.
Every Switch Needs Data the Account Earns First
Google is direct about what AI Max depends on. Its help center says search term matching, text customization and final URL expansion need a conversion-based Smart Bidding strategy to work effectively. That means a campaign bidding on bookings, with enough bookings for the bidding to learn from.
A new healthcare account doesn't have that. The rollout in this guide earns that history one campaign at a time. First the exact-match searches prove themselves, then a broad-match campaign with a cost-per-booking target joins once a service line books roughly 30 patients a month, and only after that does Performance Max come in. By the time AI Max comes up, the account already has a broad-match campaign that's learning from real patients. AI Max's job at that point is to find what Exploration missed, not to replace it.
So AI Max comes last, one service line at a time, as a test. Pick the line with the most booking history, usually the one that reached Exploration first. Copy that line's Exploration campaign, enable the AI Max features on the copy, or run it as a Google Ads experiment against the original, and compare the cost per booked patient over several weeks. Keep it only if the bookings hold up. For those first weeks, somebody opens the search terms report each morning and adds negatives before lunch, just as Chapter 6 asked for Exploration. And make sure the line isn't short on budget first. A campaign that keeps hitting its daily cap gives AI Max nowhere to go, and Google's own help page warns about exactly that.
Text Customization Reads Every Page on the Practice's Website
The switch that surprises practices most is text customization. It doesn't just rephrase the headlines from Chapter 8. Google draws on your whole domain: service pages, blog posts, old announcements, provider bios, the FAQ page no one has updated in three years.
For most businesses, the risk is an outdated price. For a medical practice, the list is longer:
- An old promotion still sitting on a page, such as a seasonal discount that ended years ago.
- Prescription drug names on a weight loss or men's health page. Chapter 10 covered how Google treats drug terms as a sensitive category.
- Outcome promises from a testimonial, such as "pain-free after one visit," that nobody would approve as a headline.
- Scary language from a health education page, which drifts into what Google calls imposing negativity.
- Condition words that make an ad sound like it knows something about the person reading it.
None of these would make it past a person reviewing ad copy. Text customization doesn't have that person. Before any AI Max test, read the site the way the machine will, and fix or remove what shouldn't become an ad. In practice, that means opening every page a crawler can reach, not just the service pages: old blog posts, event pages, provider bios, press releases, the patient forms section. Update or remove anything out of date, soften any claim you wouldn't put in a headline, and move drug names off pages that don't need them. If the site can't be cleaned up before the test, leave text customization off and test search term matching alone.
Once it's running, review what it writes. The ad's asset details show the headlines and descriptions Google generated, and in the first weeks someone should read them at least weekly, adding an exclusion or a restriction each time a line appears that the practice wouldn't have written.
Each Service Line Gets Its Own 25 Exclusions and 40 Restrictions
Google gives advertisers a way to steer text customization, called text guidelines. There are two kinds, and both are set campaign by campaign:
- Term exclusions: up to 25 words or phrases that generated text must never use.
- Messaging restrictions: up to 40 plain-language instructions about what to avoid, such as a topic, a claim or a tone.
Google adds two details worth knowing. The guidelines also apply to generated text that's already running, not just to new text. And a term exclusion can't stop prices: excluding "{{BODY}}quot; won't prevent a generated line with a dollar amount in it. For that, Google says to use a messaging restriction instead, and to include an example of what to avoid, such as "Don't mention prices, for example $99 or $1,250." One more: term exclusions only work in the language you write them in. A practice running ads in Spanish, as Chapters 8 and 9 suggested for Spanish-speaking patients, needs its exclusions written in Spanish too.
For a multi-specialty practice, the per-campaign limit is good news. A single shared campaign would have to fit pain management, weight loss and primary care into one list of 25 words. Chapter 7 gave every service line its own campaign, so every line gets its own 25 exclusions and 40 restrictions, written for its own risks.
A pain management campaign's list looks nothing like a primary care campaign's. Pain management needs to keep drug names, specific procedures the practice doesn't offer and promises of relief out of its ads. A weight loss line needs medication names and before-and-after claims out. Primary care might need little beyond a restriction on prices and on promising that every insurance plan is accepted. Start each list with the risks you found on the site, then add to it as you review what AI Max actually writes. Write restrictions the way Google recommends, with an example inside each one: "Don't promise results, for example 'pain-free' or 'guaranteed relief,'" or "Don't name prescription medications, for example the drug names on our weight loss page." A restriction that shows the machine what you mean works better than one that only says what you mean.
Final URL Expansion Needs the Same Fences as Performance Max
The third switch sends people to the page Google picks. Google says it only sends traffic to pages that match the search and fit the ad group's theme. That still leaves room for mistakes on a medical site: a patient portal login, a pay-my-bill page, a careers page, a new-patient form built for people who already booked, or an online prescribing page for a telehealth service. None of those is a place a new patient should land from an ad, and some of them, like a prescribing page, carry their own policy questions.
Chapter 10 set up exclusions for exactly these pages in Performance Max. AI Max needs the same list. Google also notes that when final URL expansion is on, pinned headlines from Chapter 8 aren't respected, so a headline you pinned for a reason may appear anywhere, or not at all. It also undoes some of the work from Chapter 4, where every ad group got the one page that answers its search. A search for a knee specialist that lands on the hip page, or a colonoscopy search that lands on a general GI overview, costs the same click and books fewer patients.
If the exclusion list isn't finished, leave final URL expansion off while testing the other switches. Nothing about AI Max requires all three at once. Testing one switch at a time is slower, but it's also the only way to know which one helped and which one hurt.
Our specialty guides cover AI Max in their own fields. The dermatology guide covers keeping "PRP" and "guaranteed" out of a dermatology account. The plastic surgery guide covers AI Max for a plastic surgery practice. The urgent care guide covers the word "emergency" and state rules on what an urgent care can call itself.
Next: One Practice, Several Businesses to Google
This guide has kept saying "one campaign per service line." Chapter 14 makes the full case: why a multi-specialty practice is several businesses as far as Google Ads is concerned, and how to split the account without splitting the practice.
Catch up on Chapter 12 on which specialties Google lets into its pay-per-lead ads, or see every chapter on the guide home for Google Ads for Healthcare. When a practice hires us for Google Ads management, AI Max waits until every line it touches has earned it.
