This post is for hospital marketing leads and PPC managers who suspect that the way patients phrase searches is changing, and want a way to adapt campaign structure without rewriting everything. It is a framework, not a set of results. It does not claim that AI Mode has changed any specific account's performance, because the reporting to prove that does not exist in Google Ads today.
The starting point is behaviour. In AI Mode, people write a situation, not a keyword. A parent may write that their child has had a fever for three days and ask which tests are done and whether a paediatrician should see them first. A person comparing options may ask for the difference between two procedures, in the same message as a budget constraint. The words are longer, the intent is mixed and the language may not be English.
Your existing structure is likely built around short, commercial phrases: a treatment, a city, a department. The framework below shows how to keep that structure while making room for the longer questions, and where to stop.
What AI Mode changes, and what is still unknown
AI Mode reached India in English in June 2025, according to Google's India announcement, and TechCrunch reported that Bengali, Kannada, Malayalam, Marathi, Tamil, Telugu and Urdu were added in October 2025. India is also a large base for voice and visual search, so conversational input is not a niche.
What is not known publicly is how ads in AI Mode behave for healthcare in India. Google's help page on ads in AI Overviews lists healthcare among the sensitive verticals excluded from placement inside the overview, and a reported May 2026 test of healthcare ads in AI Mode was described as US-only and English-only. I found no confirmation that Indian hospital ads are eligible inside AI Mode.
That should shape your expectations. This post is not about winning an AI Mode slot. It is about the fact that the same conversational phrasing shows up in ordinary Google search too, through voice, through Lens and through habit, and you can see it in your search terms report. Design for that behaviour first.
Step 1: Sort search terms by structure, not only by topic
Take your last 90 days of search terms that triggered impressions and filter for those that are long. A workable cut-off is anything of six or more words, but pick whatever separates the top tail of your own data. Add a column for each of these features and mark them yes or no:
- Contains a symptom or situation (for example, a duration or a person: "my mother", "since last week").
- Contains a comparison ("vs", "or", "difference between").
- Contains a constraint (budget, insurance, location, timing, language).
- Contains a question word (how, what, whether, which).
- Contains a named provider or place.
Do not put patient-identifying content in a shared sheet. Search terms are aggregated, but if any term contains a name or phone number, remove it before sharing.
This gives you a taxonomy of query types that is independent of department. A comparison query about two treatments and a constraint query about cost can belong to one department yet need different ad copy and different pages.
Step 2: Map each type to an ad group pattern
Most accounts will end up with four patterns.
| Query type | What the patient wants | Ad group approach | Landing page |
|---|---|---|---|
| Symptom or situation | Reassurance and who to see | Ad text about assessment and which specialist, no diagnosis | Condition or symptom page, clinically reviewed |
| Comparison | A decision aid | Neutral headlines naming the options | A comparison page, not a booking form |
| Constraint (cost, insurance, timing) | Whether this is feasible | Headlines that address the constraint truthfully | Page with fee ranges, insurance info or appointment process |
| Provider or place | A specific doctor or hospital | Brand and named-doctor ad groups | Doctor or location page |
The point is to match the ad to the question, not the keyword. If the patient asks about cost and the ad talks about excellence, the click is wasted. Notice also that two of these four do not suit a booking form. A comparison query has not reached the booking stage.
For hospital accounts that already organise campaigns by department, add these patterns inside each department rather than creating a parallel structure. The article on multi-department campaign architecture covers the department layer, and the complete guide to healthcare PPC covers the basics assumed here.
Step 3: Use broad matching deliberately, with guardrails
Long conversational queries rarely match a keyword list. Google's AI Max for Search offers a way to reach them, and its help page on optimizing campaigns with AI Max explains the three controls: search term matching, text customization and final URL expansion. Each can be set separately, and final URL expansion depends on text customization.
For a healthcare advertiser these settings need a policy before they need a budget.
- Search term matching can send your ad to queries you never listed. Start in one ad group per department, not account-wide.
- Text customization lets Google generate headlines and descriptions. In a regulated setting, ask whether a generated headline can imply an outcome, a superiority claim or a diagnosis. If the answer is yes in any case, keep it off or review generated text before scale-up.
- Final URL expansion can send the click to a page you did not choose. Google's page notes that pinned responsive search ad assets may not be used if a more relevant URL is chosen. Use URL exclusions for pages that should never receive paid traffic, such as clinician bios with outdated information or pages awaiting clinical review.
Also keep your negative keyword list active. Conversational queries often include words like "free", "home remedy" or "without a doctor", which signal the wrong audience.
Step 4: Read the report that does exist
Google does not label a search term as having come from AI Mode. What you can read is the AI Max reporting. Google's help page on AI Max reporting describes views that show search terms and landing pages matched by AI Max, and a segment on search term match type. Use these to see which long queries came through broad matching and which pages they were paired with.
Review weekly for the first month:
- Which long queries converted into your primary goal, and which only produced clicks?
- Which queries were sent to a page that does not fit the question?
- Which queries need to become negatives?
- Which recurring queries deserve their own ad group and page?
When a conversational pattern appears four or five times with different wording, promote it. Write an ad group for it with copy that answers the question in the headline, a landing page that does the same and a negative list that protects it.




