AI Search · GEO
Generative engine optimization
for practices patients ask about.
A growing share of patients now describe their situation to an assistant and ask who to see, and the assistant names two or three practices. Generative engine optimization is the work of being one of them. It is not SEO with a new label — the unit that gets returned is a passage, the competition is a summary rather than a list of ten links, and there is no position to report.
The problem
Your practice can rank well and still be invisible to the system the patient actually asked.
The patient journey that used to start with a search and ten blue links increasingly starts with a question typed in full sentences, and ends with an answer that names a handful of options. If your practice is not in that answer, the ranking underneath it did not matter, because the patient never scrolled to it.
This breaks the familiar reporting in a specific way: there is no position. A generative answer either cites you or it does not, and it may cite you for one phrasing of a question and not for a near-identical one. A practice can hold page-one rankings across its procedure set and be named in none of the answers patients are actually receiving, and nothing on a standard SEO dashboard would show it.
What decides the outcome is different too. A generative system is assembling an answer from passages it can extract, attribute and trust. That rewards pages that state a claim plainly in the first sentence rather than building to it, content with a clear author and real credentials behind it, and facts that agree with each other across the open web. It punishes vagueness, unattributed assertion, and a practice whose own details contradict themselves across directories.
And there is a failure mode particular to medicine. These systems are noticeably more conservative on health questions, which means they lean harder on signals of genuine clinical authority — board certification, society membership, a named physician attached to the claim. A practice that has all of that but has never made it machine-legible is passed over in favour of one that has.
What the work is
Five layers that decide whether you get cited.
Generative engine optimization is mostly unglamorous structural work. These are the five layers that determine whether an assistant can find you, understand you, trust you and name you.
Crawler access, checked rather than assumed
GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended and Bingbot each have their own access rules, and a practice site can be perfectly open to Google while quietly refusing three of them — usually through a security plugin or a CDN rule nobody remembers enabling. We audit what each crawler actually receives, and publish an llms.txt that gives them a clean map of the site rather than leaving them to infer one.
Passage-level answers
The unit a generative system returns is a passage, not a page. So the answer to a question goes in the first two sentences under a heading that matches how a patient would actually phrase it, and the supporting detail follows. This is close to the opposite of the build-up-then-reveal structure most practice content uses, and rewriting for it is the single highest-yield change on most sites. It also happens to make the page better for a human who is skimming at eleven at night.
Entity and author structure
A named physician, connected to their credentials, connected to the procedures they perform, expressed in structured data rather than in prose. Board certification, fellowship, society membership and publications made explicit. This is what lets a system distinguish your practice from a directory listing with the same name, and on health questions it is the signal that carries the most weight.
Consistency across the open web
Generative systems cross-check. Where your name, address, phone, procedure list or physician roster disagree between your site, your Business Profile, the medical directories and the society listings, that inconsistency reads as low confidence, and low confidence means you do not get named. Reconciling it is tedious, unglamorous and one of the most reliable improvements available.
Measurement by actual citation
We test the real assistants with the questions patients actually ask, on a schedule, and record whether your practice is named and what it is named for. That is the only honest measurement available here, because there is no rank to track. The free AI Visibility Grader runs a six-factor version of this check on any practice site.
What you get back
Citation rate, not a position you cannot see.
The reporting answers one question: when a patient asks an assistant about your procedure in your market, how often does your practice get named, and what does it get named for? Tracked across ChatGPT, Google AI Overviews and AI Mode, Gemini and Perplexity, against a fixed question set, over time.
We are deliberate about what this is worth. AI-cited traffic is still small relative to classical organic for most practices, and anyone presenting it as a replacement for search is overselling. What it is: the fastest-growing surface in patient research, one where almost no practice has done the structural work, and one where the work that wins also improves the site for ordinary search. That combination is why it is worth doing now rather than in two years.
- Citation rate across ChatGPT, AI Overviews, Gemini and Perplexity
- Per-crawler access verified, not assumed — with llms.txt published
- Physician entity and credentials expressed in structured data
- NAP, roster and procedure consistency reconciled across the open web
- The same work lifts classical organic — no separate content program
Questions
Generative engine optimization, answered plainly.
Is generative engine optimization just SEO with a new name?
They overlap and they are not the same. Both reward clear, credible, well-structured content. But GEO optimizes for extraction and citation rather than for position, the returned unit is a passage rather than a page, and there is no ranking to report — an answer either names you or it does not. A practice can hold strong rankings and be cited nowhere, which is exactly the gap this work closes.
What is the difference between GEO, AEO and AI SEO?
Mostly vocabulary. Generative engine optimization and answer engine optimization describe the same work — being the source a generated answer draws on — and AI SEO is the looser umbrella term. We use GEO because it is the phrasing the field settled on. What matters is the work, which is identical under any of the three labels.
How do you measure something with no rankings?
By asking. We run a fixed set of the questions patients actually ask, against the real assistants, on a schedule, and record whether your practice is named and in what terms. It is a sampled measurement rather than a complete one, and we describe it that way — there is no impressions report for a generative answer, and any vendor showing you one has built it from something else.
Is it too early to invest in this?
It is early, and that is most of the argument for doing it. Almost no practice has done the structural work, the work that wins also improves ordinary search performance, and the compounding asset here is credibility and consistency, which take time to establish. The honest counterpoint: AI-cited traffic is still a minority of patient research today, so this belongs alongside paid and organic rather than instead of them.
Can we start with something smaller?
Yes. The AI Visibility Grader is free and runs a six-factor check on your practice — crawler access, entity structure, passage readiness, consistency, authority signals and current citation. It takes a few minutes and tells you which of the five layers above is actually your gap, which is usually enough to decide whether this is worth a program at all.
Find out whether any assistant currently recommends your practice.
The free AI Visibility Grader checks six factors across the assistants patients actually use, and shows you what to fix first. No call required.