# 54 Healthcare AI Citation Statistics for 2026

> When patients ask an assistant which provider to use, both engines cite providers' own pages. Beside them, Google AI Mode reads Forbes's rankings and Reddit, ChatGPT reads government pages, and every service line has its own list of sites.

By Vlad Shvets, Partner @ Qvery · Published 2026-09-22 · Updated 2026-11-09 · 10 min read
Canonical: https://qvery.ai/blog/healthcare-ai-citations-statistics/

Most patient intake forms still end on the same question: how did you hear about us? A friend, your doctor, Google, your insurer's directory, other.

That box was written for referrals that had a name. More patients now put the choice to an assistant first: a third of US adults (**32%**) turned to an AI chatbot for health information in the past year, the health research group KFF found, and **41% of them** wanted to look something up before deciding whether to see a provider at all[^1].

When that patient ticks "Google" or "other", the form can't tell you which assistant they asked, or whose pages it cited, the pages an answer points to as its sources, before it named you or someone else.

> 📌 **Highlights** 
> - ChatGPT cited a government or university site in **43.33%** of its answers to patients choosing a provider; Google AI Mode in **15.00%**.
> - Forbes appeared in **26.67%** of Google AI Mode's provider-choice answers and in under 1% of ChatGPT's.
> - Reddit or YouTube: **18.33%** of Google AI Mode's provider-choice answers, none of ChatGPT's.
> - Amazon's pharmacy pages were cited in **58.04%** of online pharmacy answers, and healthcare.gov in **42.14%** of health plan answers.
> - Health questions cite public institutions: a government or university site in **74.67%** of those answers, nih.gov alone in **23.03%**.
> - The sites that cover most provider-choice answers reach **44.8%** of the health-question answers with a source; the health-question list reaches **15.05%** of the provider-choice ones.

## Patients Ask an Assistant Before They Decide Whether to Book

Clinicians still lead: Pew found that **85%** of Americans get health information from providers at least sometimes, against **22%** from AI chatbots, including 7% often or extremely often[^2].

**Most Americans get health information from providers, one in five from chatbots** (% of US adults, at least sometimes)

85% of Americans get health information from providers at least sometimes, 66% from people with a similar health issue, 36% from social media and 22% from AI chatbots.

| Label | Value |
|---|---:|
| Health care providers | 85% |
| People with a similar health issue | 66% |
| Social media | 36% |
| AI chatbots | 22% |

Share of US adults who get health information from each source at least sometimes. Pew Research Center survey of 5,111 adults, October 20 to 26, 2025. Source: Pew Research Center: Where Do Americans Get Health Information, and What Do They Trust? (2026)

The chatbot figure is the newest source on that list, and three more surveys put a size on it. They asked different questions of different people, so read them together, not as a trend line.

- KFF, February and March 2026: **32%** of adults used an AI chatbot for health information in the past year; **41%** of them wanted to look something up before deciding whether to see a provider[^1].
- Pew, June 2026: **34%** of Americans use AI chatbots for at least one of eight health reasons, and **15%** use them to help decide whether to see a doctor[^3].
- Rock Health, December 2025: **23%** of respondents had used ChatGPT for health information, against **5%** who had used a chatbot their provider offered and 4% one their insurer offered[^4].

## Both Engines Read Providers' Own Pages

Our research: the questions patients ask when choosing a telehealth service, a health plan, online therapy, an online pharmacy, a fitness or nutrition coach, or a dental or vision provider, plus general health questions, asked of ChatGPT and Google AI Mode, Google's chat-style answers, on September 21, 2026, mostly in the US and some in the UK, Canada and Australia. A citation tells you which page an answer pointed to, not how the answer used it or whether the patient booked anyone.

Among the fifteen most-cited sites for provider choice, more than half on each engine belonged to providers or plans.

Providers among the fifty most-cited sites appeared in at least **42.59%** of ChatGPT's provider-choice answers, the answers to questions like "which telehealth service should I use", and **51.67%** of Google AI Mode's. Smaller providers further down the list make these lower bounds.

## Beyond Your Pages, the Two Engines Read Different Sites

Past the providers, the two lists barely overlap: of each engine's five most-cited sites, Teladoc's is the only one on both.

**The five sites each engine cited most when patients chose a provider, in order, share of its answers**

| Google AI Mode | ChatGPT |
|---|---|
| forbes.com: 26.67% | amazon.com: 12.22% |
| goodrx.com: 15.56% | healthline.com: 10.00% |
| teladochealth.com: 13.33% | mdlive.com: 9.07% |
| reddit.com: 12.22% | healthcare.gov: 8.52% |
| telehealth.com: 11.67% | teladochealth.com: 7.96% |

Both engines read a ranking, just not the same one. Forbes appeared in **26.67%** of Google AI Mode's provider-choice answers and in under 1% of ChatGPT's. Healthline went the other way: **10.00%** of ChatGPT's answers, 7.22% of Google AI Mode's.

Google AI Mode leaned on Forbes Health's lists of the best online pharmacies and online therapy and Forbes Advisor's best health insurance companies; ChatGPT leaned on Healthline's best telehealth companies and its Talkspace against BetterHelp comparison.

Forbes Health's online therapy ranking says it compares 39 platforms on 25 data points such as cost, session length and insurance eligibility, and says that while it earns a commission from partner links, commissions don't affect its editors' evaluations[^5].

Most of those inputs (cost, session length, insurance) are facts a provider publishes; the ratings are not. If your pricing page says one thing and a ranking says another, both can sit under the same answer, and the patient has no easy way to tell which is current.

**Both engines cite providers' pages; the rest splits by engine** (% of provider-choice answers)

A provider's own site appeared in at least 42.59% of ChatGPT's provider-choice answers and 51.67% of Google AI Mode's; ChatGPT cited a government or university site in 43.33% against 15.00%, while Google AI Mode cited Forbes or Healthline in 32.22% and Reddit or YouTube in 18.33%.

| Label | ChatGPT | Google AI Mode |
|---|---:|---:|
| A provider's own site | 42.59% | 51.67% |
| Government or university site | 43.33% | 15.00% |
| Forbes or Healthline | 10.19% | 32.22% |
| Reddit or YouTube | 0.00% | 18.33% |

Share of each engine's answers to provider-choice questions citing at least one site of each kind. Our research, questions asked September 21, 2026. Source: Qvery research: patient questions on ChatGPT and Google AI Mode (September 2026)

Reddit or YouTube showed up in **18.33%** of Google AI Mode's provider-choice answers and in none of ChatGPT's. The threads under those answers included people asking each other for affordable online therapy and how to shop for a plan.

ChatGPT cited a government or university site in **43.33%** of its provider-choice answers, against **15.00%** for Google AI Mode. Those were sites like healthcare.gov's pages on plan categories and total costs, medicare.gov and hhs.gov.

On ChatGPT your plan or price page sits next to the government's explanation of how plans and costs work. If the two disagree, the patient sees both.

On September 19, The Associated Press published a story about telehealth companies exposing customers' medical data[^6]. Two days later ChatGPT cited it in **20.18%** of its telehealth provider answers, including "most trusted virtual care services". An August check could not have caught it.

Neither engine says why it reads what it reads, and our data can't tell you either. What the data does show is that a health company checking one engine sees only part of what its patients are told.

## Every Service Line Has Its Own List of Sites

Within healthcare, the sites change with what the patient is shopping for. Split the provider-choice questions by service line, each kind of service a health business sells, and the sites change almost completely.

**The three sites cited most for each service line, share of that line's answers**

| Service line | Most-cited sites |
|---|---|
| Online pharmacy | amazon.com: 58.04%  
cvs.com: 33.04%  
walgreens.com: 28.57% |
| Health plans | healthcare.gov: 42.14%  
medicare.gov: 14.29%  
irs.gov: 12.14% |
| Telehealth | teladochealth.com: 42.11%  
mdlive.com: 35.53%  
healthline.com: 27.63% |
| Online therapy | betterhelp.com: 22.30%  
nhs.uk: 18.92%  
talkspace.com: 17.57% |

These figures pool both engines and lean toward ChatGPT's answers, so read them as the shelf a patient mostly meets on ChatGPT.

- Online pharmacies are compared against Amazon's pharmacy pages, including its pricing page.
- Health plans appear beside government explanations, including IRS rules on health savings accounts and taxes.
- Online therapy services appear beside private competitors and, for UK questions, the National Health Service's talking-therapies service and the British Association for Counselling and Psychotherapy's register of counselors.

Fitness coaching and dental care had too few answers for a percentage.

The fitness answers were headed by app makers' own sites, the dental ones by insurers.

> 📌 At Qvery, we help health businesses spend less staff time keeping their prices, plans and services right across the pages assistants read, with custom AI agents that compare what each answer says about you with your own fact sheet and draft the corrections your team approves.

Your patients ask about your states, plans, and conditions. Track those questions on both engines to find the pages that matter to your service.

GEO Agent: https://qvery.ai/geo-agent/

## Health Questions Are Answered From a Library You Don't Own

Questions about deductibles or therapy effectiveness come before provider choice. They draw on a different library.

**Health questions are answered by the NIH, Mayo Clinic and the NHS** (% of health-question answers citing the site)

nih.gov was cited in 23.03% of answers to health questions, ahead of Mayo Clinic at 17.76% and the NHS at 14.47%.

| Label | Value |
|---|---:|
| nih.gov | 23.03% |
| mayoclinic.org | 17.76% |
| nhs.uk | 14.47% |
| medlineplus.gov | 12.83% |
| healthcare.gov | 11.84% |
| cdc.gov | 10.53% |
| clevelandclinic.org | 10.20% |

Share of answers to general health questions citing each site, ChatGPT and Google AI Mode together. Our research, questions asked September 21, 2026. Source: Qvery research: patient questions on ChatGPT and Google AI Mode (September 2026)

A government or university site appeared in **74.67%** of the answers to health questions, against **36.25%** of the answers to provider-choice questions. Not one of the telehealth services, pharmacies, therapy services or insurers that fill the provider-choice lists made the fifteen sites cited most for health questions. On Google AI Mode, where we have too few of these answers for a percentage, Mayo Clinic, Cleveland Clinic and YouTube led instead.

Three sites (nih.gov, nhs.uk, and healthcare.gov) cover half of the health-question answers with citations; eight cover four in five. Provider-choice answers need eight and 34, respectively.

**A handful of sites answers health questions; provider choice spreads across dozens** (sites needed)

Three sites cover half of the cited health-question answers and eight cover four in five, while provider-choice answers need eight and 34.

| Label | Health questions | Provider-choice questions |
|---|---:|---:|
| Half of cited answers | 3 sites | 8 sites |
| Four in five of cited answers | 8 sites | 34 sites |

The fewest sites that together cover half, and four in five, of the answers that cited a source, by question type. Our research, September 21, 2026. Source: Qvery research: patient questions on ChatGPT and Google AI Mode (September 2026)

So a watch list built from one kind of question misses most of the other. The sites that cover four in five of the provider-choice answers reach **44.8%** of the health-question answers that cited a source; the health-question list reaches **15.05%** of the provider-choice ones.

What you can do is make your own condition and coverage pages agree with the pages that answer the question, and link to them. Your pages come into the conversation at the provider question, and the health question often comes first.

Each service line is read beside different pages, and health questions beside a library you don't own. So choose which lines to watch first, and keep the facts on those lines' pages current.

Have Qvery Build the Monthly Check of What Assistants Say About Your Services: https://calendly.com/qvery-vlad/30min

## Frequently asked questions



### How many patients ask an AI chatbot about their health?

About a third: 32% of US adults used one for health information in the past year in KFF's survey[^1], and 34% in Pew's for at least one of eight health reasons[^3]. Among KFF's users, 41% wanted to look something up before deciding whether to see a provider.

### Which websites do ChatGPT and Google AI Mode cite for health questions?

Public health institutions led our September 2026 research: nih.gov, Mayo Clinic, and the NHS headed the list for general health questions. Provider-choice answers drew first on providers' own pages, so keep those two types of question separate.

### Does ChatGPT cite a provider's own website when patients ask which provider to use?

Yes, pricing, delivery, FAQ, and insurance pages appeared alongside providers' home pages in our research. Check the pages cited for your own services before deciding which ones to revise.

### Why does Google AI Mode cite Forbes and Reddit more often than ChatGPT?

Our research shows the difference but cannot explain why the engines choose those sources. Google AI Mode cited Forbes and patient discussions more often, while ChatGPT leaned more on government pages and Healthline.

### Do Reddit threads matter for health companies?

On Google AI Mode they do: Reddit or YouTube sat under 18.33% of its provider-choice answers, in threads where people asked each other for affordable online therapy and how to shop for a plan. ChatGPT cited neither.

### Should our company post in the patient threads it finds?

I would keep company accounts out of patient discussions comparing services. Read those threads to understand how people describe providers, and leave the conversation to the patients.

### Does a citation mean the assistant recommended us?

A citation identifies a page the answer points to; it does not establish a recommendation or a booking. Check separately whether the answer names your company and describes its services correctly.

### Can a health company get into the rankings assistants cite?

The rankings compare facts you publish: Forbes Health's therapy ranking scores cost, session length and insurance eligibility, and says it earns commissions from partner links[^5]. Ask each publisher how it chooses and updates its list, and keep the facts it compares current on your own pages.

### Do assistants cite news stories about health companies?

They can, and quickly: two days after The Associated Press published a story on telehealth companies exposing customers' medical data[^6], ChatGPT cited it in 20.18% of its telehealth provider answers.

### How often should we check which pages assistants cite?

Monthly, and again after any change to your prices or plans: a new page can enter the answers within days.

### What is a service line?

Each kind of service a health business sells: telehealth visits, online therapy, prescriptions, a health plan. Each is read beside different sites, so each needs its own list of pages to watch.

### Should each service line have its own citation list?

Yes, because patients choosing therapy encounter different sources from patients choosing a pharmacy or health plan. Give each line an owner who can check the cited answers against its current services and accepted plans.

### What should go in our fact sheet, the one list every page must agree with?

Record visit prices with and without insurance, accepted plans, states served, and the services and conditions you treat. Each row needs its source system and a person responsible for keeping it current.

### Who should approve corrections to healthcare information?

The person who owns your published facts, usually in operations or patient access, should decide what changes. Your team can edit its own pages or ask a publisher for a correction through that publisher's process.

## Sources

[^1]: KFF, "Poll: 1 in 3 Adults Are Turning to AI Chatbots for Health Information, Equaling the Share Who Use Social Media for Health" (2026), https://www.kff.org/health-information-trust/poll-1-in-3-adults-are-turning-to-ai-chatbots-for-health-information-equaling-the-share-who-use-social-media-for-health/
[^2]: Pew Research Center, "Where Do Americans Get Health Information, and What Do They Trust?" (2026), https://www.pewresearch.org/wp-content/uploads/sites/20/2026/04/PS_2026.4.7_health-information_REPORT.pdf
[^3]: Pew Research Center, "From Diagnoses to Treatments, Why Americans Use AI Chatbots for Health" (2026), https://www.pewresearch.org/science/2026/08/25/from-diagnoses-to-treatments-why-americans-use-ai-chatbots-for-health/
[^4]: Rock Health, "The tortoise and the hare of care: Health AI insights from Rock Health's 2025 Consumer Adoption Survey" (2026), https://rockhealth.com/insights/the-tortoise-and-the-hare-of-care-health-ai-insights-from-rock-healths-2025-consumer-adoption-survey/
[^5]: Forbes Health, "Best Online Therapy Platforms Of 2026" (2026), https://www.forbes.com/health/mind/best-online-therapy/
[^6]: AP News, "Telehealth companies keep exposing their customers' medical data. What should they do?" (2026), https://apnews.com/article/telehealth-prescription-glp1-ftc-hims-e6821ac2ff88d76bbfe71c08ad21cc3a