Notes
How to Rank in ChatGPT and Perplexity: How Clinics Get Cited by AI Assistants
AI assistants name sources they can lift a clean answer from, attribute to a real entity, and confirm somewhere else. For a clinic that means a direct answer at the top of the page, structured data saying who stands behind it, and the same description of you across the web. Ranking first on Google is neither necessary nor enough.
Key takeaways
- There are two pipelines: live retrieval, which you can influence this quarter, and model memory, which moves on the timescale of training runs.
- The biggest single lever is a 40 to 55 word answer directly under the heading. Assistants lift self-contained passages.
- An assistant will not name a source it cannot identify. Person or Organization schema with consistent wording is an hour of work on a template.
- Volume is small and intent is high. On my own site, AI assistants sent 2.6% of sessions and 27% of key events over 90 days.
- Default analytics filters miss part of it. Perplexity arrives in GA4 with no medium set.
Contents
- How do you rank in ChatGPT?
- What 90 days of AI traffic looks like on a consultant's site
- Six things that decide whether you are cited
- Entity clarity, the part most sites skip
- llms.txt, schema and extractability
- How to measure it without a dashboard
- What it costs, by route
- In-house, SEO agency, consultant, or wait
- How I do it
- Ask the assistants about yourself today
A patient asks ChatGPT which clinics in their city do full-arch implants and what it costs. The answer names three practices. If you are not one of them, no Google ranking recovers that conversation. It never reached a results page.
The mechanics differ from SEO and are less mysterious than the phrase "generative engine optimization" suggests. Assistants want text they can lift with confidence and a source they can name without embarrassment.
How do you rank in ChatGPT?
You rank in ChatGPT by being a source it can quote and name. Put a complete answer in the first 55 words under a question-shaped heading, mark up who wrote the page and what they do, keep that description identical across your profiles, and let the AI crawlers in. There is no position one. You are either named or you are not.
Two different pipelines sit behind that, and it helps to know which one you are trying to enter.
Retrieval answers, used by Perplexity, ChatGPT search, Gemini and Google's AI Overviews, run a live query, pull a handful of pages and compose an answer from passages they extract. Being cited here is mostly a document problem. Can the passage be lifted cleanly, and does the source look credible?
Answers from model memory have no live retrieval. There you compete on how often and how consistently you have been described across the material the model was trained on. That is a slower game, won by being mentioned in places that get scraped: directories, industry publications, your own consistently worded pages.
The first pipeline is the one you can influence this quarter.
What 90 days of AI traffic looks like on a consultant's site
I run the same playbook on my own site that I run for clinics, so here are my own numbers. They are small, and that is the point.
Over the 90 days to mid-September 2026, AI assistants sent eight sessions out of 306. Those eight sessions produced three of the eleven key events on the site, meaning a click on "Book a call" or a contact action. Three enquiries followed: two from people who found me through ChatGPT and one through Gemini, the last arriving as a structured written brief for a two-account engagement.
Two details are worth more than the totals.
Every one of those visitors landed on the home page or the About page. None came in through an article. The assistant had already answered their question. What they wanted from my site was to check that the person it named was real.
And all three were looking for the same thing, in nearly the same words: one person, not an agency. That phrase is in my schema, my profiles and my page copy, worded the same way each time. I do not think that is a coincidence, though three enquiries is not a study.
The lesson for a clinic: an assistant recommends you for the thing you are described as, consistently, everywhere. "Full-arch implants, fixed price, one surgeon" is recommendable. "Comprehensive dental care for the whole family" is not, because every practice says it.
Six things that decide whether you are cited
| Factor | What it means concretely |
|---|---|
| Direct answer up top | A 40 to 55 word answer to the page's question before any preamble. A page that buries the answer in paragraph six offers nothing liftable. |
| Question-shaped headings | H2s phrased the way a patient asks, each answered in its first sentence. Not the third. |
| Tables and lists | Structured content is quoted more often because the edges of the fact are clear. Real table markup, never a screenshot. |
| Entity clarity | Schema that says who wrote this, what they do and where else they exist. Without it your page is an anonymous document. |
| Corroboration | The same claims about you, worded consistently, in more than one place a crawler can reach. |
| Crawler access | GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot and Google-Extended allowed in robots.txt. Many sites blocked them by accident. |
Notice what is missing: keyword density, word count and backlinks. Backlinks still matter indirectly, because they influence which pages get retrieved. They are not the mechanism.
Entity clarity, the part most sites skip
An assistant will not cite a source it cannot identify.
Most clinic sites present as a collection of pages with no stated author, no stated specialization and no link to anything that confirms either. A model reading that has a passage it likes and nobody to attribute it to. So it reaches for a source that is easier to name, often a directory or a competitor with a clearer footprint.
Fixing this is an hour of work on a template. Person or Organization schema with knowsAbout and sameAs pointing at your real profiles. Article with a named author on every page that argues something. FAQPage where you have real questions on the page. The same clinic name and the same specialization wording everywhere, including Google Business Profile, directories and social profiles.
The goal is that a model resolves your clinic to one entity with a definite specialization, and not to scattered mentions that may or may not be the same business.
llms.txt, schema and extractability
llms.txt is a plain-text index at your site root listing your important pages with a one-line description of each. It is a convention and not a standard, and no assistant is obliged to read it.
Publish it anyway. It takes twenty minutes, it makes your structure legible in a single fetch, and it has no downside. Keep the descriptions factual. I go into the file format and the markup in llms.txt, schema and extractability.
Extractability is the wider point. Content rendered only by JavaScript, key facts trapped in images, prices stated as "call for pricing": each is a fact a model cannot lift, and it will find one it can lift somewhere else.
How to measure it without a dashboard
Ask the assistants directly, on a schedule, and log the answers.
- Write 20 to 30 questions a real patient would ask, in their words, including the ones where you expect to lose.
- Run them monthly across ChatGPT, Perplexity, Gemini and Google AI Overviews. Record which sources are named.
- Check server or CDN logs for GPTBot, OAI-SearchBot, ClaudeBot and PerplexityBot: whether they arrive, what they fetch, how often.
- In GA4, build a custom channel for AI assistants using the session source, matching chatgpt, gemini, perplexity, claude and copilot. Do not filter on medium. On my property Perplexity arrives with the medium not set, so a medium filter loses it.
- Add one question to your booking form: "How did you find us?" with the assistants listed by name. Patients will tell you.
- Watch Search Console for queries where you appear and clicks do not follow. That gap is often an AI Overview answering on your behalf.
The tools that automate part of this are compared in AI search monitoring tools. For now most of it is manual.
Verified example: for a luxury US healthcare clinic, the site went from effectively zero AI visibility to appearing across more than 600 AI Overview queries within six months, built on direct-answer structure, entity schema and consistent corroboration, not on new backlinks.
What it costs, by route
This market is new and quotes vary wildly. Figures below are from 2026 pricing guides, checked in September 2026.
| Route | Typical cost | Time to first citations | What it depends on |
|---|---|---|---|
| Your own team restructures the top 10 pages | Staff hours only | Weeks, once pages are recrawled | Whether someone can edit templates and schema |
| Your SEO agency adds it to the retainer | Published 2026 guides put one-time AI visibility audits at $500 to $2,000 for a light review and $2,500 to $10,000 for a deep one. Ongoing programs most often run $1,500 to $10,000 a month | Weeks to months | Whether they treat it as separate work or a relabelled SEO report |
| I do it as a fixed project | Paid audit $500, credited toward setup. The restructuring is part of setup from $1,500 | Weeks to a few months on retrieval assistants, once pages are recrawled | Number of page templates, state of existing schema, profile consistency |
| Wait | Nothing now | Never | Competitors who are named first tend to stay named |
In-house, SEO agency, consultant, or wait
| Option | Cost | Time to result | Risk | When it makes sense |
|---|---|---|---|---|
| In-house | Lowest | Fast if someone owns it | Stops after the first ten pages | You have a capable marketer and a CMS you control |
| SEO agency | Part of retainer | Medium | Activity reports without citation tracking | The agency already writes answer-first content and can show citations won |
| Consultant | Fixed project | Fast | Depends on one person | You want it built once, measured properly, then handed to your team |
| Wait | None | None | Assistants form habits from what is crawlable today | Almost never |
An SEO agency is the better choice if you also need ongoing content production at volume. This work is mostly structural. Once templates, schema and profiles are right, keeping it going is light.
How I do it
- I run your core patient questions through the assistants and record who is named in your place, and from which pages.
- I rewrite the top of your ten most important pages: question heading, 40 to 55 word answer, then the detail.
- I fix the entity: schema on the templates, one description of the clinic, the same wording on every profile you control.
- I check robots.txt and the logs so the crawlers can reach what was just fixed.
- I set up the GA4 channel and the booking form question so results are counted.
- I rerun the questions monthly and report what changed.
After 30 days the structure is live and recrawled. After 60 you should see the first named citations on retrieval assistants. After 90 there is enough data in GA4 to judge whether those visitors enquire. The service itself is described under AI search for clinics, and the wider argument in what AI search means for patient acquisition.
The paid audit covers one ad channel for one location: the account, the tracking and the path from enquiry to booked patient, ending with a prioritized fix list. It is $500, paid upfront, and credited toward setup if we work together.
You work with me directly. There are no account managers and no juniors.
15 minutes. With me, not a sales rep.
Want to know where you stand in AI answers? The paid audit includes a manual AI visibility check: your real patient questions, run in the main assistants from your city, three times each. The paid audit is $500 per channel, per location, paid upfront, and credited toward setup if we work together. You work with me directly, with no account managers in between. Book a call or write first.
Ask the assistants about yourself today
Open ChatGPT, Gemini and Perplexity and ask the question your best patient would ask. Note who is named. Then open your own top page and read the first 55 words aloud. If they do not answer the question, you know where to start. If you would like it done properly and measured, start with the audit.
Related: AI visibility trackers for clinics: what they actually measure, and how to check it yourself
Frequently asked questions
Do I need to rank on Google to be cited by ChatGPT?
No, though it helps. Retrieval answers pull from a search index, so visibility improves your odds of being fetched. The citation decision is made on how liftable and attributable your passage is, which is why pages outside the top ten are cited regularly.
Should I block AI crawlers to protect my content?
For a clinic, almost never. Blocking GPTBot or PerplexityBot removes you from answers patients are already asking for. The calculation differs for publishers who sell content. It rarely differs for a business selling a service.
Is llms.txt worth doing?
Yes, with realistic expectations. No assistant is required to read it, but it takes twenty minutes, makes your structure legible in one fetch and carries no downside. Treat it as housekeeping.
How long does it take to start appearing?
Weeks to a few months for retrieval assistants, because they query live indexes and pick up changes once pages are recrawled. Answers from model memory move with training updates, which is far slower and largely outside your control.
Does this replace SEO?
No. It shares infrastructure with SEO: crawlability, structure, authority. A page written to be cited reads differently from a page written to rank, with shorter answers, earlier, and the reasoning after.
How much traffic should I expect?
Little, and that is fine. On my own site assistants sent under 3% of sessions and over a quarter of the actions that matter. Judge it on enquiries, not visits.
Who does the work?
I do: the question set, the page rewrites, the schema, the analytics. Nothing is handed to a junior or subcontracted.
How do I verify results myself?
Keep the question log, check the GA4 AI channel for sessions and key events, and read the "How did you find us?" answers from real patients. If all three move, it is working.