A prospective patient researching a rhinoplasty in 2026 increasingly does not start with Google. They start with ChatGPT, Perplexity, or Google's own AI Overview pane, and they ask in natural language: "who is the best rhinoplasty surgeon in Scottsdale" or "what should I expect from a deep plane facelift recovery." The model returns a short, confident answer that names two or three specific surgeons or practices and links to a small number of supporting sources.
The surgeons being named in those answers are not necessarily the ones with the highest Google rankings, the largest ad budgets, or the most aggressive SEO agencies. They are the ones whose presence on the web most matches the signal an LLM uses to decide who to cite. This piece breaks down what that signal is, why traditional SEO does not produce it, and what a practice can do over the next 90 days to start showing up.
What an LLM is actually doing when it answers a query
When you type "best rhinoplasty surgeon in Scottsdale" into ChatGPT's web search mode, the model is not running a paid auction. It is not consulting a single ranked list. It is performing a retrieval pass across the open web, identifying sources it considers authoritative on the query, extracting passages it considers citable, and then composing a natural-language response that synthesizes those passages and surfaces a small number of named entities — surgeons, practices, hospitals — with links back.
The entire decision is upstream of ranking. The model is not asking "who ranks number one for this keyword." It is asking "who do credible sources mention when they describe excellent rhinoplasty results in Scottsdale." Those two questions have different answers, and the surgeons whose names show up in AI responses are the surgeons whose web presence answers the second one.
Practically, the model is weighting four signals: entity consistency across the web, citation density from credible third-party sources, passage-level clarity on the practice's own pages, and recency of substantive mentions. Each of those signals is engineerable. None of them are produced by traditional keyword-stuffed SEO.
Entity consistency: the single highest-leverage fix
An LLM does not treat your practice as a website. It treats it as an entity — a named thing in the world that exists across multiple documents. When the model encounters "Dr. Jane Smith, MD" across 14 sources and finds the credentials, the practice name, the city, the procedures listed, and the affiliations consistent across all 14, it assigns high confidence to that entity. When five of the 14 disagree — the address differs, the practice name is rendered as "Smith Aesthetics" in some places and "Smith Plastic Surgery" in others, the listed procedures don't match — the model's confidence in the entity collapses, and so does its willingness to name the entity in an answer.
Fixing entity consistency is unglamorous work. It is a sweep across the practice's Google Business Profile, the website's schema markup, the surgeon's Healthgrades and RealSelf profiles, the directory listings, the hospital affiliation pages, and the medical-society membership pages, normalizing every field to a single canonical version. This is not a one-time fix; it requires a quarterly audit because directory data drifts.
For practices that have never run an entity audit, this single project is typically the largest single lever for AI search visibility. We have seen surgeons go from zero presence in AI responses to consistent citation across ChatGPT and Perplexity inside 60 days, with the only intervention being a clean entity sweep.
Citation density: which third-party sources actually move the needle
LLMs do not weight all citations equally. A mention in a local newspaper, a peer-reviewed surgical journal, a major aesthetic publication (Allure, NewBeauty, Vogue), or a hospital system's referring-physician page is worth materially more to the model than a mention in a directory aggregator or a low-traffic blog network.
The citations that have shown up most consistently as influence sources in our analyses are, in rough order: peer-reviewed publications the surgeon is listed on as an author, media features in publications the LLM recognizes as authoritative for the beauty/health vertical, hospital affiliation pages, board certification verification pages (ABPS for plastic surgeons), structured podcast appearances (transcripts indexed and linked), and high-quality patient-community discussions that name the surgeon by full credentials.
The actionable version: a deliberate PR effort that places the surgeon in two or three legitimate vertical publications per year, plus consistent podcast appearances with transcripts published, will outperform any volume of low-quality directory and link-building work. The model is reading the company you keep, not the count of your backlinks.
Passage-level clarity on your own pages
When an LLM crawls a practice's own website, it does not consume the page the way a human reader does. It chunks the content into 100-400-word passages and evaluates each passage for whether it makes a clear, citable claim that can be lifted into an answer.
Pages that perform well in AI responses share specific structural traits: each section has a heading that matches a question a prospect might ask; the first sentence of each section answers the question directly without preamble; specific numbers, credentials, and procedure names are used in close proximity; and the passage does not require reading the rest of the page to be understood.
Pages that perform poorly tend to be long marketing-prose blocks where the answer to any given question is buried three paragraphs in, mixed with promotional language, and dependent on context the model has to assemble from elsewhere on the page. Rewriting service pages and procedure pages for passage-level clarity is a substantial project — typically 30-60 hours per practice — but it pays off across both traditional ranking and AI citation.
Recency: why this is now a continuous discipline
Traditional SEO operated on a slow update cycle. A well-optimized page could rank for two or three years without significant change. AI search runs much faster. LLMs are retrained or have their retrieval indexes refreshed on cycles measured in weeks. A practice that has not produced substantive new content — long-form blog posts, podcast episodes with transcripts, media coverage — in the last 90 days is materially less likely to be surfaced than a practice whose web presence has a steady recent footprint.
This is one of the reasons the authority flywheel discussed elsewhere on this site directly produces AI search visibility. The flywheel's distribution layer produces a continuous stream of new long-form content (YouTube long-form, podcast episodes), and the transcripts and show-notes from that content become high-quality recent mentions that the LLM ingests and weighs.
Practices running only paid acquisition and a static brochure site can sometimes rank in traditional Google for years. They rarely show up in AI Overviews more than once or twice. The web's velocity is now part of the visibility equation.
What does not work, despite what you may hear
Several tactics being aggressively marketed to medical practices as "AI SEO" produce no measurable result. The first is keyword-density rewriting — stuffing service pages with variant phrases the model is supposedly looking for. LLMs do not weight keyword density; they weight passage clarity and entity coherence. The second is mass-purchased schema markup that misrepresents the practice. Schema is helpful, but it has to be accurate; mismatched schema actively hurts entity confidence. The third is AI-generated blog content at scale. Most LLMs detect their own outputs and weight them downward as sources; large volumes of AI-written practice content can suppress visibility rather than enhance it.
The discipline that works is closer to a hybrid of public relations, technical SEO, and editorial discipline than it is to traditional rank-tracking. The practices that are getting cited by ChatGPT and Perplexity today are practices that have been deliberately building a coherent, well-documented, well-cited web presence for two to four quarters. There is no shortcut. There is a starting point.
A 90-day starting plan
If you are starting from zero AI search visibility today, the most efficient sequence we run with new partners is roughly this. Month one: entity audit and cleanup across Google Business Profile, directories, schema, and major medical-vertical platforms. Month two: passage-level rewrite of the top eight pages (homepage, surgeon bio, top four procedure pages, blog index, contact). Month three: launch a content cadence — one long-form piece with a transcript and one third-party guest appearance per month, sustained indefinitely.
By the end of month three, most practices are starting to appear sporadically in AI responses for branded queries and long-tail procedure queries. By month six, with the cadence sustained, they are typically appearing for competitive city-level queries. By month nine to twelve, they are routinely named in answers for queries where their name was previously absent. The compounding is real, and it has the same property as the authority flywheel: a competitor cannot copy a year of accumulated web presence with money.
Do AI search citations actually drive booked consults?
Yes, and the conversion rate from AI-surfaced traffic is materially higher than from traditional Google traffic, because the prospect arrives having already been told by an authoritative source that this surgeon is among the right answers to their question. Volume is still smaller than paid acquisition today, but it is the fastest-growing channel in this category.
How do we measure AI search visibility?
Three approaches: manual sampling (running your target queries in ChatGPT, Perplexity, and Google AI Overviews on a monthly cadence and logging mentions); structured tools (DataForSEO and similar services now offer LLM mention tracking); and inbound signal (tagging consult bookings that arrived via referral sources that suggest AI surfacing). Manual sampling is the most reliable starting point.
Can a competitor outspend us in this channel the way they can in paid?
Not on the same timescale. A competitor with a larger paid budget can outbid you on Meta and Google Ads tomorrow. They cannot outspend you on a 12-month accumulated web presence in a quarter. The barrier to entry is time, not money, which is why incumbents who start now have a durable advantage.
Is schema markup actually worth the effort?
Yes, when it is accurate and aligned with the rest of the entity. We use Physician, Person, MedicalBusiness, and FAQPage schema across practice sites with consistent values for credentials, board certification, affiliations, and procedures. Inaccurate schema is worse than no schema; coordinated, accurate schema is one of the strongest single signals you can send.
Will AI search replace Google?
It is already replacing portions of it for high-intent commercial and medical queries, including in plastic surgery. Google's own AI Overviews now occupy the top of the result page for many cosmetic procedure searches. Treating AI search as separate from SEO is no longer practical; treating it as the next evolution of SEO is.