AI readiness of Indian hospital websites: a 99-site tracker
Data collected on by Gaurav Phogat. Method, scoring rules and dataset are published below.
I checked 99 Indian hospital websites for the things AI assistants and AI search features read: llms.txt, robots.txt rules for AI crawlers, and schema markup. 32 publish an llms.txt file, but only 22 mark up their doctors with Physician schema and 15 use FAQPage markup. Blocking is rare: 72 of the 86 sites with a readable robots.txt never mention an AI crawler, and just 5 block OpenAI’s GPTBot. On a simple five point score, the median hospital site scores 2.
This article is part of the healthcare digital growth series, best read in order from the pillar.
Last updated: 11 November 2026.
Patients now ask ChatGPT, Gemini, Perplexity and Google’s AI features which hospital to visit and which doctor to see. In an earlier study on what AI assistants recommend, I saw how these tools pick and describe hospitals. This tracker looks at the other side: what hospital websites give those tools to work with.
I kept the scope tight. No opinions on design, no content scoring. Just five machine-readable signals that a hospital’s digital team controls directly, measured the same way on every site on the same day. Every hospital is anonymised and grouped by type.
Key findings
- 32% have an llms.txt file. 32 of 99 sites return a genuine plain-text llms.txt. At least 4 of those files say in their opening lines that an SEO plugin generated them.
- None of the 12 government and teaching hospitals has one. Specialty hospitals and chains lead at 10 of 18, then national chains at 5 of 12.
- Only 5% block GPTBot. 5 of 99 sites block OpenAI’s training crawler. Only 1 site blocks OpenAI’s search crawler, OAI-SearchBot.
- 72 of 86 robots.txt files ignore AI crawlers. They neither allow nor block GPTBot, ClaudeBot, PerplexityBot or Google-Extended by name, so the general rules apply.
- 54% use a medical schema type. 53 sites mark up the homepage as a Hospital, MedicalOrganization, MedicalClinic or MedicalBusiness. 75% have some structured data.
- 22% mark up doctors. Physician schema appeared on 22 sites. Where I could open a doctor profile page, 20 of 44 carried it.
- Median score: 2 out of 5. Only 3 sites scored 5. 10 scored zero, including 4 of the 5 government hospitals.
How I did this
Sample. I started from a purposive list of 108 Indian hospital websites across six types: national multi-city chains, regional chains, metro and tier 2 standalone hospitals, specialty hospitals and chains (eye, fertility, cancer, orthopaedics, cardiac and mother and child), private teaching hospitals and trusts, and government institutes. 9 sites were dropped: 7 failed to load or could not be read in the browser, 1 redirected to another domain mid-check, and 1 now redirects to a group already in the sample. The final sample is 99 sites: 12 national chains, 16 regional chains, 41 standalone hospitals, 18 specialty hospitals and chains, 7 teaching hospitals or trusts and 5 government hospitals. The sample is not random, so results describe these sites rather than every hospital in India.
Collection. All checks ran on 11 November 2026 in a desktop Chrome browser, without logging in, clicking forms or accepting anything. For each site I loaded the homepage, then fetched /robots.txt and /llms.txt from the same domain. I followed the homepage link to the doctor directory and, where the directory page linked to an individual doctor, one doctor profile page.
What counts. Homepage schema was read from the rendered page, both JSON-LD and microdata. Directory and profile pages were read as server-delivered HTML. A site counts as having llms.txt only if the file returned status 200 as plain text and was not an HTML page in disguise. I opened every one of the 33 files that passed the automated check, and dropped one that turned out to be an HTML page. For robots.txt, each crawler was matched to the rule group that names it, or to the general * group if none does. “Blocked” means a Disallow: / rule for that crawler, with or without a few Allow exceptions.
Score. One point each for: a working llms.txt; a readable robots.txt that does not block GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot or Google-Extended; a medical organisation schema type on the homepage; FAQPage markup on any page checked; and Physician markup on any page checked.
Limits. This is a one-day snapshot. Robots.txt only shows declared intent; a firewall or CDN can still block bots that robots.txt allows, and I did not test that. Schema added by JavaScript on directory and profile pages would be missed. My script found a doctor profile link on 44 sites, so the Physician figure is a floor, not a ceiling. Some sites redirected or timed out on single files; those are recorded as not readable rather than guessed.
The basics are mostly there, the detail is not
Most hospital sites have done the first layer of technical SEO. 86 of 99 serve a readable robots.txt and 74 put some structured data on the homepage. The drop comes one level down, where the markup describes what a patient actually asks about: which doctors work here, and what are the common questions.
Only 53 sites tell machines plainly that they are a hospital or medical organisation. Another 17 describe themselves only with generic types such as Organization or LocalBusiness, which is better than nothing but loses the medical context. 6 sites had at least one broken JSON-LD block that would not parse, which means that markup is invisible to any system reading it.
Doctor markup is the weak spot. Physician schema showed up on 22 sites. On the 44 sites where I could open a doctor profile, 20 had it, so the problem is partly that many directories load doctors with JavaScript and give machines no crawlable profile links at all. I found the same pattern from the human side in my study of doctor profile pages.
Hardly anyone is blocking AI crawlers
There is a lot of talk about publishers blocking AI bots. Indian hospitals mostly have not joined in. Of the 86 sites with a readable robots.txt, 72 never name an AI crawler. Their general rules apply, and in almost every case those rules leave the site open.
The 5 sites that block GPTBot did it in three different ways. One blocks every crawler from the whole site with a single rule, search engines included. Two use a managed robots.txt from their CDN, which blocks training crawlers such as GPTBot, ClaudeBot and Google-Extended while leaving OpenAI’s and Perplexity’s search crawlers open. Two others block GPTBot alone and leave ChatGPT’s search and browsing agents untouched.
That split matters. Blocking a training crawler and blocking a search or retrieval crawler are different decisions. A hospital that blocks OAI-SearchBot or PerplexityBot is asking not to be cited in those answers, and only one site in the sample made that choice, through its blanket rule.
llms.txt arrived faster than doctor schema
The surprise for me was llms.txt. It is a young, informal proposal and no major AI company has said it uses the file for ranking. Yet 32 hospital sites already publish one, more than the number that mark up their doctors. Some are long, curated site maps. At least 4 open with a line saying an SEO plugin generated them, which suggests some sites switched it on as a plugin setting rather than as a planned project.
Adoption tracks marketing maturity more than size. Specialty hospitals and chains, which compete hard for self-paying patients in categories like fertility and eye care, lead at 10 of 18. Government and teaching hospitals, which rarely run search marketing, have none. Only 1 of the 5 government sites served a readable robots.txt, and only 1 had any structured data at all.
Most sites score 1 or 2 out of 5
Put the five signals together and the picture is a long middle. 55 of the 99 sites score 1 or 2. 21 score 4 or 5. National chains and specialty hospitals have a median of 3; regional chains and standalone hospitals sit at 2; teaching hospitals at 1 and government hospitals at 0.
A score of 1 usually means the site has an open robots.txt and nothing else that helps an AI system describe it accurately. That is not a failing grade on its own. It just means the AI answer about that hospital is being assembled from directories, reviews and news coverage rather than from the hospital’s own facts. I made the same point in how hospitals get cited by AI search.
What this means for hospital marketers
These are digital and brand steps, not clinical or legal advice. The aim is to make your own facts easy for machines to read, so AI answers about your hospital quote you rather than a third party.
- Decide your AI crawler policy on purpose. Read your robots.txt and your CDN’s bot settings together. If you want to appear in ChatGPT search, Perplexity and AI Overviews, do not block OAI-SearchBot, PerplexityBot or Googlebot. Blocking training crawlers is a separate choice.
- Fix the homepage entity first. Use Hospital or MedicalOrganization, with name, address, phone, departments and a link to your Google Business Profile. Validate it so a broken block does not hide everything.
- Make doctor profiles crawlable. Give every doctor a plain HTML profile URL linked from the directory, with Physician schema covering name, specialty, qualifications, languages and hospital. Keep claims factual and in line with the NMC’s 2026 advertising guidelines.
- Add FAQ markup where the questions are real. Timings, visiting hours, insurance and TPA tie-ups, health check packages and how to book. Answer in plain language and avoid outcome claims.
- Treat llms.txt as a cheap extra, not a strategy. If you publish one, curate it: key service pages, doctor directory, locations, contact. Do not leave a plugin default pointing at old blog posts.
- Re-check after every redesign. Agency rebuilds often drop schema and reset robots.txt. A ten minute check after each release is enough.
For the wider debate on terms, see my explainer on AEO vs GEO vs AI SEO, and for the full technical baseline across 100 sites, the Indian hospital website benchmark.
Download the data
The full dataset (CSV) has one row per website in shuffled order: an anonymised ID, hospital type, robots.txt status, whether the robots.txt is CDN managed, how many AI crawlers it names, the block status for six AI crawlers, llms.txt presence, the seven schema flags, whether a doctor directory and profile page were fetched, the AI readiness score and the collection date. It contains no hospital names, domains, doctor names or cities. Collection date: 11 November 2026. The data is licensed under CC BY 4.0: you can reuse it with credit and a link to this page.
How to cite this study
Phogat, G. (2026). AI readiness of Indian hospital websites: a 99-site tracker. gauravphogat.com. https://gauravphogat.com/hospital-website-ai-readiness-tracker/ Data collected 11 November 2026. Licensed under CC BY 4.0.
Questions people ask
Sources and further reading
- CC BY 4.0 (creativecommons.org)
llms.txt is a plain-text file at the root of a website that lists the pages a site wants AI systems to read first. It is an informal proposal, and no major AI company has confirmed it affects ranking. In my check of 99 Indian hospital websites on 11 November 2026, 32 had a working llms.txt. It is cheap to add, but crawlable doctor pages and clean schema matter more.
It depends on what you want. GPTBot, ClaudeBot and Google-Extended relate mainly to model training, while OAI-SearchBot, ChatGPT-User and PerplexityBot fetch pages for search answers. Blocking the search crawlers can keep your own pages out of those answers. Only 5 of 99 hospital sites I checked block GPTBot, and only 1 blocks OAI-SearchBot.
Start with Hospital or MedicalOrganization on the homepage, then Physician on every doctor profile, then FAQPage where you answer real patient questions such as timings, insurance and booking. In my sample, 54% used a medical organisation type, 22% used Physician and 15% used FAQPage.
22 of the 99 hospital websites I checked had Physician schema on at least one page. On the 44 sites where I could open an individual doctor profile, 20 carried it. Many directories load doctors with JavaScript and offer no crawlable profile links.
Each site got one point for a working llms.txt, one for a readable robots.txt that does not block the main AI answer engine crawlers, one for a medical organisation schema type on the homepage, one for FAQPage markup and one for Physician markup. The median score across 99 sites was 2 out of 5, and 3 sites scored the full 5.
Read my takes first in Google Search
