The Multi-Location Clinic That AI Could Not Pin Down
Northline Health operates 14 clinics across three states and six service lines, including urgent care and primary care. Its new AEO initiative is prompted by a problem: when patients ask an AI assistant “Does Northline accept Medicaid?” or “Which Northline clinic is open Sunday?”, answers sometimes name the wrong location, mix up insurance policies, or cite a directory listing with outdated hours. The website has location pages, but they share about 60% duplicate copy, phone numbers vary between tracking numbers and main lines, and hours differ across the site, Yelp, and Healthgrades. There is no Organization or LocalBusiness structured data, and the brand name is sometimes written “Northline Health” and sometimes “Northline Clinic.” The team has one content strategist, one web developer, and a $25,000 budget for the next quarter. Leadership wants “AI visibility” but has not defined what a correct answer looks like.
1. What foundation work should the team prioritize in the first 30 days, and why should it come before publishing new AEO content?
Read the worked answer
The first priority is not new content; it is eliminating contradictions. Answer engines synthesize from multiple sources, so a new FAQ page will not fix a wrong clinic name or outdated Sunday hours if directories and location pages still disagree. The team should create a single canonical fact source for each location: legal name, public name, address, main phone, hours, service lines, insurance accepted, and URL. Then reconcile the website, Google Business Profile, Yelp, Healthgrades, and any other high-authority directory.
Second, standardize the brand and location naming convention. “Northline Health” versus “Northline Clinic” creates entity ambiguity, especially across 14 locations. Third, add Organization and LocalBusiness/MedicalClinic structured data using the canonical facts, and remove tracking numbers from structured data where they differ from the public main line.
This order matters because AI answers are only as good as the consistency of the evidence they retrieve. Publishing more pages into a conflicting data environment increases the chance of another wrong answer. The 30-day deliverable should be a verified facts database, corrected priority listings, and a defined “correct answer” rubric for the prompts leadership cares about.
2. How should Northline model locations and services as entities so answer engines can cite the correct clinic?
Read the worked answer
Use one canonical page per physical location, not one page per service per city. Each location page should carry the same name, address, phone, hours, and service list as the canonical database, plus a visible “last reviewed” date and a short, unique description of that clinic’s staff, access notes, and insurance caveats. Duplicate boilerplate should be reduced, not multiplied.
Model the organization and each clinic as distinct entities. Organization schema should identify the parent brand; LocalBusiness or MedicalClinic schema should identify each location and link back to the parent. Service-line pages should describe the service and link to the locations that offer it, rather than duplicating the same service copy on every clinic page. SameAs links should point to the verified Google Business Profile and authoritative directory listings.
Separate “what service is this” from “where can I get it.” That lets an answer engine cite the service page for a definition and the location page for hours or insurance. For insurance, publish a clear, dated policy page per state or clinic and link to it from the location page. Ambiguity is the enemy: the model should be able to tell which clinic, which service, and which policy applies.
3. What evidence would show the foundation is working, beyond keyword rankings?
Read the worked answer
Measure correctness, not just presence. Build a prompt set that reflects real patient questions: “Northline [city] Sunday hours,” “Does Northline accept Medicaid in [state],” “Northline urgent care vs primary care,” and “Northline [city] phone number.” Run it monthly across the assistants that matter, and score each answer as correct and cited, correct but uncited, incomplete, or wrong. The trend in wrong answers is the key foundation metric.
Also track directory consistency: the percentage of priority listings that match the canonical facts, and the number of unresolved conflicts by location. On the website, monitor calls and direction requests by location, not just sitewide, because a foundation fix should show up at the clinic level. Branded search by clinic name is another useful signal. Finally, keep a simple manual QA log with screenshots and dates so leadership can see specific wrong answers being fixed. Rankings alone cannot show whether an AI assistant told a patient the wrong insurer or the wrong Sunday hours.