A patient opens ChatGPT and types: “I need a dermatologist in Sacramento who takes Blue Shield and can see me about a mole this month.”
The assistant answers with two or three names. Not a list of ten blue links — a short, confident recommendation, sometimes with a sentence explaining why each one fits. The patient picks one, maybe checks the website, and calls.
Your practice is either in that answer or it isn’t. There is no page two.
This is a different game from search engine optimization, and the tactics that worked for Google rankings only partly overlap. Here is what actually determines whether an assistant names you.
AI assistants don’t rank — they assemble
A search engine returns a ranked list and lets the human decide. A language model does something else: it gathers information from sources it can access, forms a judgment about which options fit the question, and states an answer.
That difference has three consequences that matter enormously for a clinic.
Confidence beats position. A model will not recommend a practice it cannot describe. If it can’t determine your hours, whether you take a given insurance, or what you actually treat, it hedges — and hedging usually means naming someone else. Being findable is not enough. You have to be describable.
Corroboration matters more than authority. Traditional SEO rewarded domain authority accumulated over years. Assistants care more about whether multiple independent sources agree. Your website says you’re open Saturdays, your Google profile says you’re open Saturdays, your listing on a health directory says the same — that agreement is what turns a hedge into a recommendation.
Specificity wins. A model answering “best dermatologist for adult acne in Sacramento” is looking for a practice that visibly treats adult acne. A page that says “comprehensive dermatologic care for the whole family” matches nothing in particular and therefore matches nothing at all.
The five things that decide whether you get named
1. Structured, machine-readable practice facts
Everything a patient might filter on has to exist as text a machine can parse. Not an image of your hours. Not a PDF of accepted insurance plans. Not a JavaScript booking widget that renders availability only after a human clicks.
At minimum:
- Services and conditions treated, named specifically
- Providers with credentials, training, and specialties
- Insurance plans accepted, listed by name
- Locations, hours, and holiday hours
- Whether you accept new patients, and how quickly
Schema markup makes this explicit rather than inferred. MedicalBusiness, Physician, MedicalProcedure, and FAQPage markup tell a machine what each piece of information is, instead of leaving it to guess from prose.
2. Consistency across every source
Your name, address, phone number, and hours need to be identical everywhere they appear — your site, Google Business Profile, Apple Maps, Bing Places, health directories, insurance provider finders, and review platforms.
Inconsistency is uniquely damaging here. A model that finds two different phone numbers does not average them. It loses confidence in everything it knows about you, and confidence is exactly what determines whether it recommends you.
This is unglamorous work and it is usually the single highest-return thing an independent practice can fix.
3. Question-shaped content
The unit an assistant lifts into an answer is a self-contained explanation. Not a paragraph that only makes sense after the two above it. Not a sentence fragment inside a marketing block.
Write pages that answer the actual questions patients ask, and answer them completely in place:
- What does a dental implant cost, and what changes the price?
- Do I need a referral for a sleep study?
- How long is recovery from a rhinoplasty, realistically?
- Does this practice take my vision plan?
Practices consistently underestimate how much cost questions matter. Price is among the first things patients ask an assistant, and practices that answer honestly — with ranges and the variables that move them — get cited far more than practices that write “contact us for pricing.”
4. Third-party corroboration
Models weight independent sources heavily, because they are harder to manipulate than your own website. Reviews on Google and specialty platforms, listings in professional directories, hospital affiliations, society memberships, local press, and mentions on referring practices’ sites all function as verification.
You cannot fabricate this, and you shouldn’t try. What you can do is make sure the legitimate corroboration you already have is complete, current, and consistent with everything else.
5. Freshness
Citations decay. A practice that was accurate eighteen months ago and has since changed hours, added a provider, or dropped an insurance plan becomes a liability for a model to recommend. Assistants are conservative about medical recommendations, and stale information reads as risk.
This is the part most practices get wrong. Visibility is not a project you complete. It is a state you maintain.
What doesn’t work
You cannot buy a recommendation. There is no ad placement inside a ChatGPT answer for local services today. That is genuinely good news for independent practices — a well-structured solo clinic can outrank a badly-structured group with ten times the budget.
Keyword stuffing actively hurts. Models are reading for meaning. Repetition without substance makes a page harder to summarize, not easier.
Mass-generated content backfires. Fifty thin pages about fifty procedures you barely perform will dilute what a model understands about you. Ten genuinely useful pages about what you actually do will outperform them.
Optimizing for one assistant leaves most of the demand untouched. ChatGPT, Gemini, Claude, Perplexity, and Google AI Overviews weight sources differently. Coverage matters more than perfecting your standing with any one of them.
How to know whether it’s working
Ask the questions your patients ask. Run them against each major assistant on a regular schedule. Record whether you appear, who appears instead, and how the answer is phrased.
That’s it. There is no ranking dashboard for AI citations the way there is for search position, so measurement means systematically checking the actual output. Do it monthly at minimum. Do it for the questions that would fill your schedule, not vanity queries about your practice name — of course an assistant can find you when someone types your name. The question is what happens when they don’t know it yet.
A realistic timeline
Listing and profile corrections propagate within days. Assistants pick up changes on their own crawl schedules, which vary from near-real-time for engines that search live to weeks for models relying on training snapshots and cached indexes.
Meaningful movement in citations typically takes six to twelve weeks. Practices in less competitive metros move faster. Practices in saturated markets — aesthetics in Los Angeles, dental in Phoenix — take longer and need more genuine differentiation to work with.
The window is unusually open right now. Most independent practices have done none of this. That will not stay true.
Trendzz builds and maintains this footprint continuously for clinics — structured practice facts, listing consistency, question-shaped content, and citation tracking across every major assistant. See how it works.