Making an Orthopedic Practice Visible to AI

I have spent the last several years in operating rooms as a spine implant rep. I have watched surgeons build practices, and I have watched where their cases actually come from. So when I started looking at how specialty practices show up in AI search, I was looking at a problem I already understood from the inside.

Here is what changed, and most practices have not adjusted for it.

Patients used to arrive at a consult knowing they had back pain. Now they arrive having asked an assistant "what is the difference between fusion and disc replacement," "how long is recovery from a laminectomy," "should I get a second opinion before spine surgery." They arrive with a vocabulary and a short list. Adult children researching for a parent do the same thing, in more depth.

The assistant gave them that vocabulary. It also, in many cases, gave them names. Rarely a solo practice, because solo practice websites are built for nobody in particular and cannot be read, verified, or cited.

I built this concept to work out what fixing that actually requires.

What does the practice look like today?

Kestrel is a mid-sized orthopedic group with a spine division. Multiple fellowship-trained surgeons, hospital affiliations, an ASC relationship, and a website that is a brochure: a photo of the building, a services list, a phone number, and a stock image of an anatomical spine model.

It is not a bad website. It is an invisible one.

Why is health content different?

This vertical is not like the others, and pretending otherwise is how these projects fail.

Health content sits in the highest-scrutiny category there is. Assistants are deliberately conservative about it. They hedge, they refuse to give individual medical advice, and they default to citing large institutional sources over any individual practice. You are not going to out-rank the Cleveland Clinic on "what causes sciatica," and chasing that is a waste of a budget.

So the strategy inverts. You do not compete for the medical explanation. You compete for everything the institutions cannot answer, which is everything local, practical, and specific to a surgeon:

• Who does this procedure in this metro area, and what is their training
• What does the recovery timeline actually look like week by week here
• Which insurance is accepted, and what does prior authorization involve
• What happens at a second opinion visit and how do you get records sent
• What is the difference between having this done at a hospital versus an ambulatory surgery center

Institutional sources are generic by design. That gap is the entire opportunity, and it is where patients actually get stuck.

What does the research actually show?

I did not want to guess at this, so here is what the published work says.

A study analyzing 5,472 citations produced by ChatGPT, Claude, Gemini, and Perplexity in response to health prompts found Cleveland Clinic cited 174 times and Mayo Clinic 163 times, with roughly 30% of all citations coming from health media sites and about 23% from commercial or professional practice pages.

A January 2026 analysis of ChatGPT health responses found the same shape: citations came primarily from established institutional sources, with under a quarter drawn from commercial and professional practice websites.

A 2026 healthcare citation index put Mayo Clinic as the most-cited source in the entire healthcare category, ahead of every pharma brand and consumer health platform. The stated mechanism is not brand recognition. It is publishing physician-reviewed consumer health content at scale.

Two conclusions follow, and they set the whole strategy.

The ceiling is real. Individual practices are a minority of citations and will stay that way. Anyone promising you the top spot on a clinical explainer query is selling you a campaign you will lose.

The floor is not zero. Roughly a quarter of citations do go to commercial and professional practice pages. That share is winnable, and it is almost entirely uncontested, because practices are not competing for it.

The lever is the physician-reviewed byline. That is what separates the sources that get cited from the sources that do not, and it is available to a ten-surgeon group just as much as to a hospital system.

What do the sites that get cited actually do?

I did not want to take the published research on faith either, so I audited twelve orthopedic sites in three groups: the institutions assistants actually cite (Mayo Clinic, Cleveland Clinic, Hospital for Special Surgery, and the AAOS patient education site OrthoInfo), four large private orthopedic groups with real marketing budgets, and four solo or small-group surgeon sites. Raw HTML, not rendered pages. Three patterns separate the winners from everyone else.

The first sentence under every heading is the answer. On all four institutional sites, every section I sampled opens with a self-contained definition. Cleveland Clinic answers "What is a laminectomy?" with "A laminectomy is surgery on your spine to remove the lamina." An assistant can lift that section and it stands alone. The private practices open with taglines instead. One large group's knee page starts with "our nationally recognized physicians bring advanced training." That sentence answers nothing, so nothing on the page gets quoted.

A named human signs the content. HSS is the model: articles carry a visible "medically reviewed by" line naming a surgeon and a date, and the markup underneath models that surgeon as a full entity with specialty, job title, and a linked profile. Across roughly thirty pages of large private group sites I found zero bylines and zero review lines. Not weak ones. None. One practice's markup tells machines that every clinical page, including the surgeon bios, was authored by its website vendor. These groups employ fellowship-trained surgeons from Mayo and Hopkins and publish everything as an anonymous brand voice.

The surgeon exists as one consistent entity, or not at all. I traced individual surgeons from their own sites to hospital directories to review platforms. The typical private-practice surgeon appears in three to five name formats across surfaces, sometimes with conflicting cities and addresses. Assistants resolve answers to entities. A surgeon whose identity fragments across the web is a surgeon an assistant cannot confidently recommend.

And even the citation winners leave ground uncovered. None of the twelve sites publishes a week-by-week recovery table. None publishes procedure costs. Only one institution has a real second-opinion page, and only one private group in twelve lists its accepted insurance as plain text on a page. The practical layer patients actually ask about is nearly empty at every tier.

What happened when I asked the assistants?

On August 20, 2026 I asked ChatGPT, Claude, Gemini, and Perplexity the same question: who does robotic knee replacement in Bergen County, New Jersey. All four named the same solo surgeon in the answer body, on the same day. Not a hospital system with a nine-figure budget. One surgeon, because his website states his robot platforms, his locations, and his differentiator in plain checkable sentences. Perplexity put him in the first row of its answer table, above both major hospital systems. Claude's search returned his site in three of its twenty sources. Gemini and ChatGPT both reproduced claims from his pages nearly verbatim.

Every other joint surgeon in the county was absent from all four answers. Their sites are not worse doctors' sites. They are unquotable, so the patient asking an assistant never hears their names.

The same pattern held on the practical queries. On recovery timelines, private practices that published specific numbers dominated the source lists. On an insurance-constrained local query, the assistants returned named surgeons sourced from practice websites. The clinical explainers belong to the institutions. The commercial queries are open, and mostly uncontested.

These results carry a date because they move. Answers shift with every model release, which is exactly why visibility is a monthly measurement and not a one-time fix.

Why does this matter commercially?

The demand is already there and it is upstream of the consult.

A KFF tracking poll conducted in early 2026 found 32% of US adults had used AI tools or chatbots for health information in the previous year. Among those users, 41% said a major reason was to look up information before deciding whether to see a provider.

That is the moment the practice is currently absent from. Not the moment a patient chooses between two surgeons, but the earlier one, where they decide whether this is worth an appointment at all.

What makes a practice citable?

Every claim traceable to a credentialed human. Content is authored to a specific surgeon, not to the practice. Fellowship, board certification, hospital privileges, and NPI attached to that person as a structured entity. Where a clinical claim is made, a named physician reviewed it and the page says so with a date. In this vertical, unattributed content is worse than no content.

Structured data that matches reality. Physician and MedicalClinic markup, procedure pages marked up as procedures, FAQ markup on every question and answer pair, hospital affiliations and accepted insurance as machine-readable fields rather than a PDF.

Consistency across independent surfaces. Practice name, each surgeon's name and credentials, and the address formatted identically on the site, hospital directory pages, payer directories, professional society listings, and review platforms. Confidence in AI systems comes from agreement across sources. Specialty practices are usually a mess here, with three spellings of the same doctor across five directories.

Direct answers, then depth. Question as a heading, plain answer in the first two sentences, then nuance, then a citable figure with a linked primary source. No introductions about how the practice is committed to excellence.

Honest scope. Every page that touches a clinical decision says plainly that it is educational and not a substitute for evaluation. This is not just liability hygiene, it is a trust signal that both patients and models respond to.

What about the referral channel?

Consumer search is only part of where spine volume comes from. Primary care, physical therapy, pain management, and chiropractic referrals still drive a large share of it, and referring providers do their own looking up.

So the site carries a second track built for them: a clear referral pathway, what this surgeon takes and what they send back, typical time to consult, and how to send imaging. Most practice sites have nothing for the referrer, which is strange given who actually fills the schedule.

How would you measure whether it works?

Not rankings. Consult requests attributed to the site, procedure page engagement depth, referral form submissions, and a monthly check on whether the practice gets named when assistants are asked the local and procedural queries in the content plan.

That last number moves with every model release, so it is a standing monthly measurement rather than a one-time optimization.