An llms.txt file for a healthcare website is a short, plain-text guide placed at the root of your site that tells AI systems which of your pages are worth reading. For a clinic, hospital, or health publisher, it should list your best patient-education pages, condition and treatment guides, service pages, provider bios, and location pages — and it should never list anything that touches patient data. It takes about an hour to write, costs nothing to host, and gives AI answer engines a clean map of the health content you actually want them to find.
That is the short version. The longer version matters more for health sites than for almost any other kind of site, because health content sits in the category Google calls "Your Money or Your Life" — pages where wrong information can cause real harm. AI systems that answer health questions lean heavily on sources they can trust, and a careless llms.txt file can point them at the wrong pages, or worse, at pages that should never have been public in the first place. This guide walks through exactly what to include, what to leave out, and how llms.txt works alongside robots.txt.
What llms.txt actually is
llms.txt is a proposed convention, not a web standard. The idea comes from llmstxt.org: a Markdown file at https://yoursite.com/llms.txt that lists a site's most important pages with a short description of each, so large language models can find the good material without crawling every page. An optional companion file, llms-full.txt, holds the full text of those pages in one document.
Think of it as a table of contents written for machines. A human visitor uses your navigation menu; an AI system uses your llms.txt. The file itself is simple: a title, a short summary, then sections of links with one-line descriptions. There is no special software to install and no fee. Any site can publish one today.
One honest note before we go further: llms.txt is still a proposal. Google's John Mueller has said publicly that no AI system currently uses llms.txt, and as of late 2026 there is no confirmed evidence that the major answer engines read the file as a matter of routine. Some smaller AI tools and crawlers do fetch it. That makes llms.txt a cheap bet, not a guaranteed channel. You should write one because it costs almost nothing and may help, not because it will transform your traffic overnight. Everything in this guide is written with that limit in mind.
Why healthcare sites need to be more careful than most
Most sites can afford a sloppy llms.txt. A health site cannot. Three reasons:
First, health content is held to a higher standard. Google's search quality guidelines treat health pages as YMYL content, and its guidance on creating helpful content stresses first-hand expertise and trust. AI answer engines face the same pressure: when they answer a question about symptoms, medication, or treatment options, they need sources written or reviewed by qualified people. Your llms.txt should steer them toward the pages where your expertise is strongest, not toward every page you happen to have.
Second, a mistake can expose private information. llms.txt is a public file. Anything you list in it is an invitation to fetch that URL. If a URL leads to a patient portal, a test-results page, or a form that pre-fills personal details, listing it is a privacy failure. The rule is simple: if a page needs a login, or if it shows anything about an identifiable patient, it does not belong in llms.txt — or in any public index.
Third, outdated health information is dangerous. A treatment page from 2021 that still recommends a withdrawn drug, or a condition guide that predates new clinical guidance, can mislead an AI answer and, through it, a real patient. Your llms.txt should only list pages you have reviewed recently and plan to keep current. If a page is out of date, fix it first or leave it out.
What to list in a healthcare llms.txt
Start with the pages a patient or caregiver would actually want an AI assistant to find. For most clinics and hospitals, that means five groups of pages.
1. Condition and treatment guides
These are your strongest candidates. A well-written page on type 2 diabetes, knee replacement recovery, or childhood asthma — written in plain language, reviewed by a clinician, and kept up to date — is exactly the kind of page an AI answer engine wants to cite. List the guides you are proudest of first, and keep the descriptions specific: "What to expect in the first six weeks after a total knee replacement, including exercises and warning signs" is useful; "Knee page" is not.
2. Patient education and self-care pages
Pages that explain how to prepare for a procedure, how to take a medication safely, or when to seek urgent care are high-value for AI answers, because those are the questions people actually ask voice assistants and chatbots. Include them, and note the review date in the description if you can, for example: "Reviewed by our cardiology team, March 2026."
3. Services and departments
List your main service pages — cardiology, orthopedics, pediatrics, urgent care, telehealth, imaging — with a one-line description of what each service covers and who it is for. This helps AI systems answer "where can I get an MRI near me?" or "does this clinic offer telehealth?" with your real pages instead of a directory listing.
4. Provider bios
Provider pages do real work for trust. A bio that lists a doctor's qualifications, specialties, and languages spoken helps an AI answer establish that a human expert stands behind the content. List the bios of clinicians who author or review your content, and link each bio to the articles they reviewed where that makes sense.
5. Locations and practical information
Include your location pages, opening hours, contact details, insurance and billing information, and accessibility notes. These are the pages people need when they move from asking a question to booking an appointment, and they are often the pages AI systems get wrong when they rely on third-party directories.
What to leave out — the non-negotiable list
What you exclude matters at least as much as what you include. Never list:
- Patient portals and login-only pages. Appointment booking systems, test-result portals, messaging tools, and any URL behind authentication. These pages are not public content, and listing them invites crawlers to probe them.
- Anything containing patient data. That includes case studies with identifiable details, testimonial pages that name patients, and before-and-after galleries. Even de-identified stories deserve caution — if a page could be traced back to a real person, leave it out.
- Internal staff pages. Intranet links, staff directories with personal contact details, rota pages, and admin areas have no place in a public file.
- Outdated or unreviewed health content. If a page has not been reviewed by a clinician in the last year or two, or if it covers a topic where guidance has changed, keep it out until it has been updated. An AI system citing your old page is worse than it citing nothing.
- Thin or duplicate pages. Tag archives, search-result pages, printer-friendly versions, and near-duplicate location pages add noise. Your llms.txt should be a short list of your best work, not a copy of your sitemap.
A useful test: before adding any URL, open it in a private browser window. If you would be uncomfortable with a stranger reading it, or with an AI system quoting it in an answer to a patient, do not list it.
A worked example: llms.txt for a fictional clinic
Here is a complete example for a fictional multi-specialty clinic, Riverside Family Health. Notice the structure: a title, a short summary, then sections with links and one-line descriptions. The file is plain Markdown, and every link is a real page on the site.
# Riverside Family Health > Riverside Family Health is a multi-specialty clinic in Austin, Texas, offering primary care, cardiology, pediatrics, orthopedics, and telehealth. Our patient education pages are written by our clinical team and reviewed annually. ## Patient education - [Managing type 2 diabetes: a patient guide](https://riversidefamilyhealth.example.com/patient-education/type-2-diabetes-guide): Everyday blood sugar management, diet, exercise, and when to call your care team. Reviewed by our endocrinology team, February 2026. - [Preparing for a knee replacement](https://riversidefamilyhealth.example.com/patient-education/knee-replacement-preparation): What to expect before, during, and after surgery, including home preparation and recovery exercises. - [Childhood asthma: a guide for parents](https://riversidefamilyhealth.example.com/patient-education/childhood-asthma-parents): Recognizing symptoms, using inhalers correctly, and creating an asthma action plan. ## Conditions and treatments - [High blood pressure (hypertension)](https://riversidefamilyhealth.example.com/conditions/hypertension): Causes, symptoms, diagnosis, and treatment options, including lifestyle changes and medication. - [Knee osteoarthritis treatments](https://riversidefamilyhealth.example.com/conditions/knee-osteoarthritis): Non-surgical and surgical options, from physical therapy to joint replacement. ## Services - [Cardiology](https://riversidefamilyhealth.example.com/services/cardiology): Heart health screening, echocardiograms, and ongoing care for heart conditions. - [Telehealth appointments](https://riversidefamilyhealth.example.com/services/telehealth): How virtual visits work, what you need, and which conditions can be treated online. - [Urgent care](https://riversidefamilyhealth.example.com/services/urgent-care): Walk-in care for minor injuries and illnesses, hours, and what to bring. ## Our clinicians - [Dr. Maya Chen, cardiology](https://riversidefamilyhealth.example.com/providers/dr-maya-chen): Board-certified cardiologist specializing in preventive cardiology and heart failure. - [Dr. James Okafor, pediatrics](https://riversidefamilyhealth.example.com/providers/dr-james-okafor): Pediatrician with 15 years of experience in childhood asthma and allergies. ## Locations and practical information - [Austin central clinic](https://riversidefamilyhealth.example.com/locations/austin-central): Address, opening hours, parking, and accessibility. - [Insurance and billing](https://riversidefamilyhealth.example.com/patients/insurance-billing): Accepted insurance plans, self-pay options, and billing questions. ## Optional - [Full patient education library](https://riversidefamilyhealth.example.com/llms-full.txt): The complete text of our patient education pages in a single file.
A few things to notice. The summary line states who the clinic is and that its pages are clinician-reviewed — that is the trust signal in one sentence. Each description says what the page covers and who it helps, not just its title. The Optional section points to an llms-full.txt file for tools that want the whole text at once. And nothing in the file requires a login or reveals anything about a patient.
How to write yours, step by step
Step 1: Audit your content first. List every health page on your site and mark each one as current, needs updating, or out of date. Only current pages go in the file. If your audit turns up pages that have not been reviewed in over a year, that is useful work on its own — fix them before you publish anything.
Step 2: Pick your 15 to 40 best pages. A good llms.txt is short. Start with the pages that answer the most common patient questions and the pages your clinicians are proudest of. You can always add more later; a tight file that an AI system can read in seconds beats a 500-link dump.
Step 3: Write one-line descriptions. Each link needs a description that says what the page covers and who it is for. Use plain language and name the condition or service. Skip marketing language — "world-class care" tells an AI system nothing.
Step 4: Add a review note. Where a page has been clinically reviewed, say so in the description with a month and year. This is a small detail that helps both AI systems and human readers judge whether the page is current. If your site has a medical review policy, link to it in the summary.
Step 5: Put the file at the root of your site. The file must live at /llms.txt — that is, https://yoursite.com/llms.txt. A file buried in a subfolder will not be found by tools that look for it. If you also create an llms-full.txt, link to it from the Optional section.
Step 6: Generate the file instead of hand-typing it. If your site has hundreds of pages, hand-writing the file invites broken links. A generator can build the file from your sitemap and check that every URL returns a working page. Our free llms.txt generator builds a draft file from your site's pages, which you can then edit down to your best content. Whichever tool you use, read the output before publishing it — a generator cannot tell which of your pages are clinically current.
Step 7: Check it and schedule a review. Open the live file in a browser, click every link, and confirm each one works. Then put a reminder in your calendar to review the file every three to six months, or whenever clinical guidance changes for a listed topic. A stale llms.txt is worse than no llms.txt.
How llms.txt pairs with robots.txt
llms.txt and robots.txt do opposite jobs, and health sites need both. llms.txt is a guide: it says "these are the pages worth reading." robots.txt is a gate: it says which crawlers may fetch which parts of your site at all.
That distinction matters because not every AI crawler is an answer engine. Some crawlers collect content to train models. A hospital may be happy for its patient-education pages to appear in AI answers but unwilling to have the same pages used as training data. robots.txt is where you draw that line, by blocking specific training crawlers while allowing search and answer-engine crawlers.
A practical setup for a clinic looks like this: use robots.txt to block AI training bots you do not want, keep your patient portal and internal systems disallowed, and use llms.txt to point the crawlers you do allow toward your best public health content. The two files work together — one controls access, the other directs attention. If you need help writing the robots.txt side, our free robots.txt generator builds a file with the major AI crawlers already listed, and the robots.txt tester checks whether a given bot is allowed or blocked on any URL. For a deeper look at how the two files fit together, see our guide to llms.txt for law firms, which covers the same pairing for another field where accuracy and trust carry real weight.
E-E-A-T and health content in AI answers
Google's quality guidelines use four words to judge content: experience, expertise, authoritativeness, and trustworthiness — E-E-A-T for short. Health content is judged more strictly than most topics because bad advice can harm people, and AI systems that answer health questions apply the same logic when choosing sources.
Your llms.txt can support E-E-A-T in three concrete ways. First, list pages that name their author or reviewer — a page reviewed by a named clinician with stated credentials is a stronger source than an anonymous page. Second, keep the file short and honest: listing only your reviewed, current pages tells an AI system that you curate your content, which is itself a trust signal. Third, make sure the pages you list actually demonstrate expertise — real clinical detail, clear sourcing, and honest statements about what is not yet known beat generic summaries every time.
One thing llms.txt cannot do is create trust you do not have. If your health pages are thin, anonymous, or out of date, no file will make an AI system cite them. Fix the content first; the file comes second. That is also why the file should never list pages you would not want a patient to act on. An AI answer that quotes your outdated dosage page is a patient-safety problem, not just an SEO problem.
The honest limits of llms.txt for health sites
It is worth stating the limits plainly, because health sites are often sold expensive services on the back of new standards. llms.txt is a proposal from 2024, it has no official status with Google or any major AI company, and no major system has confirmed that it reads the file. Google's John Mueller has said no AI system currently uses it. Some smaller AI tools and research crawlers do fetch it, and the file is cheap to publish, so the cost-benefit still favors writing one — but it should be an hour of work, not a project, and not a line item you pay an agency to maintain.
What does move the needle for health content in AI answers is the same work that has always mattered: accurate, clinician-reviewed pages, clear authorship, fast and accessible pages, and genuine expertise that other sites link to. llms.txt is a small signpost on top of that work. Write it, keep it current, and then get back to the content itself.
For a sense of how the evidence for llms.txt is shaping up, our study of whether AI bots actually read llms.txt looks at real server logs, and our guide to llms.txt for local businesses covers the same file for clinics and practices that also need patients to find their locations. Google's own guidance on creating helpful, reliable, people-first content is the best single reference for the standard your listed pages need to meet.
FAQ
Does llms.txt help a healthcare website rank on Google?
There is no evidence that Google uses llms.txt for ranking, and Google's John Mueller has said no AI system currently uses the file. Its value is for AI tools and crawlers that do fetch it, and as a clean, curated map of your best content. Treat it as a small, cheap experiment, not an SEO strategy — your rankings will still come from the quality of your pages.
Can I list pages that require a login, like a patient portal?
No. llms.txt is a public file, and anything listed in it is an invitation for a crawler to fetch that URL. Patient portals, test-result pages, appointment systems, and anything behind authentication must stay out. The same goes for any page that shows information about an identifiable patient.
Does llms.txt create any HIPAA risk?
The file itself is not a HIPAA problem, as long as it only lists public pages. The risk comes from listing the wrong URLs — a portal, a form that pre-fills patient details, or a page with identifiable patient stories. Keep the file limited to public, reviewed education and service pages, and the file adds no privacy risk at all.
Should a small clinic bother with llms.txt?
Yes, if it takes less than an hour. A small practice with 20 good pages can write a useful file in one sitting, and the discipline of choosing your best pages is valuable even if no AI system reads the file today. If your pages are thin or out of date, spend the hour updating them instead and come back to llms.txt later.
How often should a healthcare llms.txt be updated?
Review it every three to six months, and immediately whenever clinical guidance changes for a listed topic — a new treatment guideline, a withdrawn drug, or updated screening recommendations. Also update it whenever you publish a major new patient guide or retire an old one. Put the review on a calendar so it does not get forgotten.
What is the difference between llms.txt and llms-full.txt?
llms.txt is a short index: links with one-line descriptions. llms-full.txt is the full text of all the listed pages in a single file, for tools that want the content itself rather than links. Most health sites should start with llms.txt only, and add llms-full.txt later if they want to — it is optional, and keeping a full-text copy in sync with your pages is real maintenance work.
