Patients do not browse dental websites the way they browse restaurant menus. Before they book an appointment, they need to know three things fast: does this practice take my insurance, are they accepting new patients, and do they handle my specific problem. Increasingly, they ask an AI assistant instead of clicking through to your website — and AI gets all three of those questions wrong more often than most practice owners realize.
The failure is not random. Healthcare AI errors follow a predictable pattern, and fixing the underlying data gaps is more straightforward than it sounds.
The questions patients actually ask
When someone opens ChatGPT or Perplexity and types a question about a dental or medical practice, the queries cluster around five topics — in roughly this order of frequency:
Insurance acceptance. "Does Riverside Family Dentistry take Delta Dental PPO?" or "Does this doctor accept Aetna?" This is the most dangerous question for AI to answer incorrectly, because the patient will discover the mistake only after they have already booked, driven in, and sat in the waiting room. At that point they are not annoyed at the AI — they are annoyed at your front desk.
New patient status. "Is Dr. Chen's office accepting new patients?" A practice that closed its books six months ago still shows as "accepting new patients" in most AI systems, because nothing in the typical data trail reflects that change. The AI answers confidently with stale information.
Emergency and same-day availability. "Does Lakeview Dental do emergency appointments?" Patients with a cracked tooth at 7 PM are making fast decisions. If AI says no — or hedges because it has no data — they call your competitor.
Specific services. "Does this dentist offer Invisalign?" "Is pediatric dentistry available here?" "Does the clinic do sleep apnea treatment?" Practices that list services only inside a PDF brochure, or buried in a sub-page with no structured markup, are essentially invisible to AI on these questions.
Payment and cost. "Does this practice offer payment plans?" "What does a new patient exam cost?" The angle here is different from a restaurant price question. Healthcare patients asking about cost are often anxious about affordability — they are not comparison-shopping for a deal. An AI that says "pricing information is unavailable" does not come across as neutral; it reads as a warning sign.
Why the stakes are higher than for other local businesses
A coffee shop that AI describes incorrectly loses a customer who can easily recover from the mistake. A medical or dental practice that AI misrepresents loses a patient who may never rebook — and who may leave a confused review about the experience. More importantly, healthcare trust is built before the first appointment. Patients decide whether to call based on how credible and transparent the practice appears. An AI answer that is vague, outdated, or wrong undermines that credibility before anyone picks up the phone.
This is worth understanding concretely. If Perplexity tells a prospective patient that Northside Dental accepts Cigna PPO and it does not, the patient does not blame Perplexity. They feel misled by the practice.
Where AI finds information about your practice
AI assistants pull from a short list of sources when answering questions about healthcare providers: your Google Business Profile, your website (especially your homepage and services pages), third-party directories like Zocdoc, Healthgrades, and WebMD, and the NPI registry for physician credentials. If those sources disagree with each other — or if key details are simply missing — the AI either guesses or hedges.
The NPI registry is worth a specific mention. It is a public federal database that AI can access, and it contains specialty codes, practice addresses, and organizational affiliations. If the address in your NPI record does not match your Google Business Profile, AI gets confused about where you are. Many practices relocated during 2020–2022 and never updated their NPI record. Check yours at nppes.cms.hhs.gov.
The specific fixes that move the needle
Google Business Profile services. GBP has a dedicated services section. List each service you actually offer — by name, not by category. "Invisalign provider," "same-day emergency appointments," "pediatric patients welcome," "payment plans available" are all indexable signals that feed directly into AI responses about your practice.
Insurance listed on the website, not in a PDF. A PDF is invisible to AI. A plain HTML page titled "Insurance We Accept" with each carrier listed in text is readable. One page, updated when your contracts change, does a significant amount of work.
Schema markup. The Dentist and MedicalOrganization schema types let you declare your accepted insurance, available services, and patient age range in structured data that AI can parse reliably. If you are already using LocalBusiness schema, the healthcare-specific types are an extension of the same pattern.
New patient status as a live signal. This one is harder, but important. If you are not accepting new patients, say so explicitly on the homepage — not buried in a contact form. AI cannot read your scheduling software, but it can read a sentence on your website.
Keep your business hours consistent everywhere. Hours on your GBP, your website, and your Healthgrades profile should match exactly. Discrepancies do not just confuse AI — they erode trust with any patient who notices the difference.
The fastest way to see where your practice currently stands is to run a free AI visibility scan — it checks the same questions a patient would ask and shows you where the gaps are.