A clinic I looked at recently publishes a solid blog post every week, and those posts earn AI citations. Its knee-pain service page? Not touched in four years. When patients asked AI about knee treatment, the model pulled from both, but the neglected service page was still carrying weight despite zero attention.
That gap is the missed opportunity in most healthcare content marketing. In Yolando's June 2026 analysis of 27,812 AI answers, the median cited brand site had a single page driving around 32% of all its citations. That page is the homepage 43% of the time, a service or condition page 12%, and a blog post 5%.
This does not mean blog content is weak. It means service pages and blog posts do different jobs. When AI cites your own domain, it usually lands on a page describing what you do. Blog content earns citations off your domain, on the third-party sites, roundups, and threads that good posts generate. Blog and content pages account for around 53% of all AI citations across engines. You need both page types, and the one most clinics have let go quiet is the service page.
What a service or condition page is retrieved for
A service or condition page gets retrieved when it answers a specific patient question an AI assistant is trying to resolve. There are seven of those questions, and a page that skips one hands that answer to whatever source does cover it.
Here are the seven questions every service or condition page has to answer:
Do you treat this? The condition or service, named plainly, in the words a patient uses.
For whom? Who is eligible, and who is not.
Where? Which locations or delivery modes (in-clinic, telehealth) apply.
How? What the treatment actually involves, step by step.
What is involved for the patient? Duration, recovery, what it feels like.
What does it cost? The pricing logic, even when you cannot publish a fixed number.
What happens next? The concrete next action to book or refer.
In rater8's 2026 Patient Choice Report, almost half of patients now use AI chatbots to help them find a doctor, and those bots sometimes return the wrong address or detail. A page that covers all seven questions is how you become the detail the bot gets right.
Skip any one and the model fills the gap from another source. Miss eligibility and a competitor answers it. Miss cost logic and a telehealth brand wins the "how much does this cost" prompt. AI hedges healthcare answers more than almost any other category: in Yolando's June 2026 analysis, hedging rates ranged from 20% in mental health to 70% in lab testing. A page that closes every question gives the model enough grounded material to skip the hedge and actually recommend you.
Provider and brand sites account for about 61% of all healthcare citations, the largest single source type. Your own pages are the one layer you fully control, and most clinics under-use it.
The template, block by block
Nine blocks make up a complete service or condition page. Each one below includes the heading pattern, what belongs in it, a rough word count, and a filled example you can copy into your CMS.
H1 pattern
Use [Service or condition] treatment in [modifier], plain, no cleverness. Your H1 is the single strongest signal of what the page is about.
Example: Knee osteoarthritis treatment for adults. Keep it under about 60 characters. Skip taglines and marketing verbs.
Opening definition, written to be liftable verbatim
Your first two or three sentences need to fully answer "what is this and do you treat it," with no throat-clearing. Write them as a self-contained paragraph a model can quote whole. Around 40 to 60 words.
Example: "Knee osteoarthritis is the gradual wearing of cartilage in the knee joint, causing pain, stiffness, and reduced mobility. Our clinic treats it with a staged plan starting with physical therapy and injections, moving to surgical referral only when conservative care stops working."
Eligibility and contraindications
Say who this is for and who it is not for. Most pages skip this one, and skipping it is what hands the answer away. Around 60 to 100 words, ideally as a two-column list.
Example: "Good candidates: adults with diagnosed knee osteoarthritis, ongoing pain despite rest, no active joint infection. Not suitable for: untreated knee infection, certain bleeding disorders, pregnancy (we adjust the plan first)."
What the service actually involves, step by step
Lay out the process as a numbered list. AI systems lift numbered steps almost verbatim, so this is one of the easiest blocks to get cited from. Around 80 to 120 words.
Example:
Initial assessment and imaging review.
A conservative plan: physical therapy, weight guidance, anti-inflammatory options.
Image-guided injections if pain persists.
Reassessment at six to eight weeks.
Surgical referral only if conservative care has been exhausted.
Evidence and outcomes
Say what the treatment achieves in honest terms and cite the evidence on the page. Never invent an outcome number. Around 60 to 90 words.
Example: "Most patients report meaningful pain reduction within two to three months of conservative care. Injection response varies. We follow treatment guidance from the American Academy of Orthopaedic Surgeons rather than promising a fixed result." Link the clinical body inline.
Credentials
Name the clinicians who deliver the service, with qualifications and registration, in a form a machine can read. Around 40 to 70 words.
Example: "Dr. Ana Reyes, MD, board-certified orthopedic surgeon (registration #12345), and Sam Okafor, DPT, lead physical therapy." Put credentials in text, not inside an image.
Practical logistics and cost logic
Answer duration, recovery, frequency, and pricing. You don't need a fixed price, but you do need the logic. Around 70 to 110 words.
Example: "First visit: 45 minutes. Injections: 20 minutes, same-day discharge. Conservative programs span 8 to 12 weeks. Cost depends on insurance and whether imaging is needed; we provide a written estimate before any procedure."
FAQ block
Five to seven questions in the exact words patients use, not clinical phrasing. Pull them from intake calls and search queries. Each answer 40 to 80 words, self-contained.
Example: "Does a knee injection hurt?" Not "What is the analgesic profile of intra-articular injection?"
Next step
One clear action, one link. No competing asks. Around 20 to 40 words.
Example: "Book a knee assessment online or call the clinic for a referral. New patients usually get an appointment within a week."
The evidence and credential slots, done properly
Source every clinical claim by name and put every credential in plain text.
Sources. Cite named clinical bodies or government health agencies. "Studies show" is not a source. Name the body and year.
Citations. Put the reference inline, next to the claim it supports, and link it where a public URL exists.
Credentials. Plain text, not inside an image or PDF. A model can read "Dr. Ana Reyes, MD, board-certified, registration #12345." It cannot read a headshot.
Outcome language. "Most patients report reduced pain within three months" is defensible. "Guaranteed pain relief" is not. AI hedges healthcare answers at rates up to 70% by vertical, and measured language is what clears that hedge.
Page-level schema and internal linking
Implement MedicalWebPage as the page type, MedicalCondition or MedicalProcedure for the subject, Physician for clinicians, and FAQPage for the question block. FAQ markup is the one most directly tied to appearing in an AI answer, so prioritize it if you only implement one.
For internal linking, link each service page to its hub and to sibling pages a patient would move between (knee to hip). If you run one page per condition per metro, the condition-and-city dermatology blueprint is the worked example.
A maintenance rule
Review every service and condition page on practice events, not on a calendar. A page untouched for three years reads as stale to systems that weight freshness, and a stale clinical page can be both inaccurate and ignored.
Four events should trigger a review:
A new clinician joins or leaves the service. Update the credential block.
A changed protocol in how you deliver the treatment. Update the steps.
New evidence or updated guidance from the body you cite. Update the evidence block and its date.
A price change or insurance change. Update the logistics block.
Three complete pages beat thirty thin ones
Build three complete pages before thirty thin ones. A single page drove around 32% of the median cited brand's citations, and 29 of the 31 most-named local mental-health clinics were cited from their own website. Pages that go deep get quoted; thin ones don't.
Keep the blog running too: it earns the third-party coverage that makes up the bulk of your citations.
Yolando's healthcare AI visibility platform shows you which pages AI cites and which questions your service pages are missing. Book an audit. No marketing team to act on the results? Yolando In A Box handles it for you.
