Which AI assistant sends the most visitors to healthcare websites in India?
AI Overview Citation Study: India Healthcare 2026
155 Indian healthcare websites · 10,787 AI-referred sessions (Jul–Sep 2026) · 2,073 pages audited
Short answer: ChatGPT. Across 155 Indian healthcare websites in ICG's 2026 study, ChatGPT sent 92.2% of all AI-assistant referrals that could be attributed (9,942 of 10,787 sessions, July to September 2026). Gemini sent 3.4%, Perplexity 2.4%, Claude 1.5% and Copilot 0.6%. Volumes are still modest, but ChatGPT referrals grew 65.6% between July and August.
Hospital and clinic marketing teams are being asked a new question: "How much traffic are we getting from AI?" The honest answer for most is "we don't know", because their analytics do not separate it. This article uses the AI-referral layer of ICG's AI Overview citation study for Indian healthcare to show where that traffic comes from, how fast it is growing, and how to measure it properly.
One assistant sends almost all of it
Share of attributable AI-assistant sessions, Indian healthcare websites
n = 10,787 sessions with an identifiable AI referrer · 155 websites · July–September 2026
The split is so lopsided that it changes how a team should prioritise. If you are going to check how one assistant describes your hospital, or which of your pages it links to, start with ChatGPT. The other assistants are worth monitoring, but in this dataset none of them sent more than one AI-referred visit in twenty-five.
One important exclusion: Google's AI Overviews appear inside Google Search, and clicks from them arrive as normal Google organic traffic. They are not in these numbers; the study looked at Overview and Gemini citations separately, and the full report covers them.
Small, but growing fast
| ChatGPT-referred traffic, 155 websites | July 2026 | August 2026 |
|---|---|---|
| Sessions | 3,284 | 5,437 |
| GA4 key events | 3,069 | 5,172 |
| Key events per session | 0.93 | 0.95 |
| Month-on-month growth in sessions | – | +65.6% |
Spread across 155 websites, 5,437 sessions a month is a small channel: a few dozen visits per site on average, far below what Google Search sends. What makes it worth tracking is the trend and the reach. In August, 97 of the 155 websites (62.6%) received at least one ChatGPT-referred visit; Gemini reached 27 and Claude 10. This is no longer something only a few specialist sites see.
These visitors also did more on arrival. In August they triggered 0.95 GA4 key events per session. Key events are whatever each site has marked as important (calls, WhatsApp clicks, form views and so on), so this is a measure of engagement, not of bookings. A plausible explanation is that people arriving from a chat have already asked their questions and come to act, but the study did not test that directly.
How to see this traffic in your own analytics
- Look before you assume it is zero. In GA4, open the traffic acquisition report by session source and search for chatgpt.com, perplexity.ai, gemini.google.com and claude.ai. Many teams find visits already there, filed under referral.
- Create an AI-assistants channel group. GA4 lets you define custom channel groups. Add one that matches the AI assistants' referrer hosts so the channel appears in every standard report, alongside organic search and paid.
- Check what strips the referrer. Consent banners, redirects and tag-manager overrides can turn an AI referral into "direct". If AI traffic is unexpectedly absent, test this first.
- Report it monthly. Sessions, key events, landing pages and the share of total traffic. Landing pages tell you which of your content AI assistants are linking to.
How to become a page AI assistants link to
No one can guarantee a citation, and anyone who promises one is overselling. But the same study audited 2,073 pages across the 155 websites, and the gaps it found are the practical starting point:
- Doctor identity. Only 13.7% of pages carried Physician schema. Doctor pages with clear names, qualifications and registration details, marked up consistently, make it easier for any system to understand who is behind the advice.
- Authorship and freshness. Only 45.1% of pages showed a named byline and 60.6% a visible updated date. Both are cheap template changes.
- Answer-first writing. Pages that answer the question in the first paragraph, then explain, are easier to quote accurately.
- Consistent facts everywhere. The same name, address, phone number, doctors and services on your website, Google Business Profile and directories.
In healthcare, all of this sits inside the rules: NMC professional-conduct limits on self-promotion, ASCI codes on claims, and the DPDP Act 2023 for any patient data on the page. Content written to be cited should be educational, not promotional.
ICG's AI search optimisation and healthcare SEO teams handle tracking and citation readiness together, alongside healthcare content and website development. For the attribution detail, see ChatGPT referral traffic attribution for healthcare marketers.
Read the full report
This article uses one finding from AI Overview Citation Study: India Healthcare 2026. The report has the full tables, the method and the limits, and it is free to quote with credit and a link.
Want your own numbers next to these? Ask for a free check of your AI-assistant traffic tracking and citation readiness. ICG works with 150+ healthcare brands; there is no obligation, and we will tell you if nothing needs fixing.
Related reading
- Doctors evaluating these assistants for their own work can start with our guide to AI tools for doctors in India.