Insights

When patients ask AI first: what a provider’s website has to answer

What does AI tell patients about your practice, and where does it look?

October 26, 2026

· Healthcare

· 12 min read

About one in three U.S. adults use AI chatbots for health, some to help decide whether to see a doctor. Google and OpenAI say their AI search answers draw on web pages, and your website is the source you control most fully. Make it the public reference for what you treat, who you take, where and when, and the next step.

Put the question a new patient would ask to an AI tool, such as ChatGPT or the AI answers in Google search. Who near me treats this, takes my plan and can see me soon? The answer arrives as a short paragraph with a few names and a link or two. Your practice may be in it, with last year’s hours and a plan you dropped in June. Or the answer names a competitor, and its link opens a page that lists plans and hours where yours says to call.

Before the next marketing dollar, the team needs two answers. What does AI say about us, and where does it look?

01

How many people ask AI about their health?

In a Pew Research Center survey in June 2026, 34 percent of U.S. adults said they use AI chatbots for health reasons. Among all U.S. adults, 15 percent said they use them to help decide whether to go to a doctor. That decision comes before anyone calls your office.

KFF polled U.S. adults in late February and early March 2026. About a third had turned to AI chatbots for health information in the past year. Both surveys asked about health in general, so the step to choosing a practice is our reading.

In our reading, search data shows a related pull toward the answer itself. In March 2025, Google users clicked a result about half as often when an AI summary appeared at the top. Searches that began with a question word, such as who or why, drew an AI summary 60 percent of the time.

Figure 1
Google users clicked a result about half as often when an AI summary appeared
Share of visits to a Google results page with a click on a traditional result link, March 2025
Bar chart. In March 2025, visits to a Google results page with traditional results only included a click on a traditional result link 15 percent of the time. Visits to a page that showed an AI summary included such a click 8 percent of the time, about half as often. A link inside the AI summary drew a click in 1 percent of the visits that showed one. About one in five searches drew an AI summary, and searches that began with who, what, when or why drew one 60 percent of the time. Data from Pew Research Center, browsing by 900 U.S. adults.
A link inside the AI summary drew a click in 1% of the visits that showed one. About one in five searches drew an AI summary; searches that began with who, what, when or why drew one 60% of the time.Source: Pew Research Center, browsing by 900 U.S. adults in March 2025.

People were also more likely to end their browsing after a page with an AI summary. We read this as a sign that the answer can be where the search ends. What it says about your practice can then matter more than the links beneath it.

02

Where does the answer come from?

From pages and listings across the web, your own site among them. How much comes from your site rests partly on whether AI tools can read it. Google’s AI answers come in two forms, AI Overviews atop some search results and AI Mode for more complex questions. Both may run several related searches, then link to pages that support the answer.

Google says the best practices for search remain relevant to those answers, starting with pages it can read and the important facts in text. It adds that new AI text files or markup are unnecessary to appear in them. We would test any pitch for special tactics in Google’s AI answers against that list.

ChatGPT’s search draws on the web too. OpenAI says its search crawler, a program that visits pages, surfaces websites in ChatGPT’s search results. It adds that a site blocking that crawler drops out of those answers, though it can still appear as a navigational link. OpenAI also runs a training crawler, which gathers pages that may be used to train its models. A site can allow or opt out of each crawler separately.

We would treat crawler access as a business decision, made with your legal and privacy teams. OpenAI adds that visits ChatGPT makes when a user asks a question may fall outside a site’s crawler settings. A public page stays readable to any visitor, so we would treat these settings as choices about visibility.

Google says the rules for its search crawler manage crawling for Search, and its AI answers are part of Search. It adds that crawler access is set in several places, from the site to its hosting. Have your web team check each one against the decision.

We would expect the answer to lean on other listings when your site leaves a fact out, and those listings can carry errors. A study checked Marketplace plan directories by phone in seven states and six specialties, from May 2022 to December 2023. In 14.5 percent of listings checked, the directory wrongly said a provider was taking new patients. Each of those can send your front desk a call it has to turn away. The risk, as we read it, is that an answer drawn from listings like these can repeat their errors.

03

What does the answer need from your site?

The opening question asks for four facts. They are what you treat, who you take, where and when, and the next step. What you treat belongs in the words patients use. How your site describes a condition or treatment is for your clinicians and compliance team to decide. This piece stays with the four access facts.

Who you take means plans by name, as your contracts list them, with the date they were last checked. Add a line that asks patients to confirm coverage with their plan. Where and when means each location, its services, hours and any telehealth. The next step means new-patient status, any referral your practice or the patient’s plan asks for, and the way to book.

Some sites keep these facts apart from the page. The plan list sits in a scanned PDF, the hours in an image, and a line says to call. Google recommends keeping important content in text.

Our piece on referral sources argued that these facts need an owner in operations. In our view, AI is now a second reader of the same page, so give each fact an owner who already holds it. Clinical leaders own what you treat, contracting owns the plan list, each location’s manager owns hours and services, and scheduling owns new-patient status. When contracting sends a health plan an update, the site gets it too. Marketing publishes each change to the site and the listings you manage.

Google’s own list names the Business Profile, the listing it shows in Search and Maps. Keep the same facts there and in every directory listing you can edit.

Figure 2
Four facts in text on your site, each with an owner
Each fact with the date last checked, and the same on your site and every listing you manage.
Fact 1
What you treat
OwnerClinical leaders, with compliance on the wording
Where it hidesA services menu written for clinicians
On the siteConditions and services in patients’ words, as your clinicians approve them
Fact 2
Who you take
OwnerContracting
Where it hidesA plan list in a scanned PDF, or a line that says to call
On the sitePlans by name, the date last checked and a prompt to confirm coverage with the plan
Fact 3
Where and when
OwnerEach location’s manager
Where it hidesHours in an image
On the siteEach location, with its services, hours and any telehealth in text
Fact 4
The next step
OwnerScheduling
Where it hidesA phone number as the only way in
On the siteNew-patient status, any referral your practice or the plan asks for, and the way to book
Source: jscTEK.

We would fix these four facts before funding the next campaign, because every campaign sends patients to them.

04

How do you check what AI tells patients?

Give one person a fixed list of questions and a log. Each month, that person puts them to ChatGPT and Google’s AI Mode for each location, by town, covering services, plans, hours and booking. Keep the questions general, and use the AI tools your organization allows. Pew notes that an AI summary may change over time, so we would treat each check as a sample.

Compare each answer with your site, correct what you control, and send the rest to the listing’s owner. Changes to how you describe care go through your clinicians. After any change to plans, hours or locations, update every listing you control, then ask again.

Then measure what arrives, with the analytics tools your privacy team has approved. ChatGPT marks the visits it sends from its search results with a tag, so your analytics team can count the tagged visits. Google counts clicks from its AI answers in the overall search traffic it reports in Search Console, within the Web search type. We read that to mean those clicks sit inside your search totals, unlabeled.

Both measures cover only the patients who click through. We would watch another number more closely, the new-patient calls turned away over plan, location or availability.

05

What should the team decide this quarter?

Three decisions, in order.

  1. Make the site the public reference for the four facts. Give each fact its owner, from clinical leaders to scheduling, and show the date it was last checked.
  2. Decide which AI crawlers your site settings allow, with your legal and privacy teams. Then have your web team confirm that every site, hosting and security setting matches the decision.
  3. Check what AI tells patients, every month. Score each answer by location, correct what you control and track the new-patient calls you turn away.

On Monday, ask the team three questions. Ask an AI tool which plans you take. Is the answer right? Where did it come from? When did someone last check that page?

If that page lacks an owner, it may still carry last year’s answer, and an AI answer can repeat it.

The first step

The first step is a conversation about what your site tells new patients. Bring the kinds of questions they ask, in general terms and free of any patient’s details. We put those questions to AI tools with you and look for the page or listing behind each answer. Then we look at where your site states the four facts, and who keeps each one current.

At jscTEK, we understand how healthcare buyers decide. See what we build on our Work page.

Sources

  1. How many adults use AI chatbots for health, and why. Emma Kikuchi, Giancarlo Pasquini and Eileen Yam, Pew Research Center, “From Diagnoses to Treatments, Why Americans Use AI Chatbots for Health,” August 25, 2026, a survey of 3,488 U.S. adults, June 22 to 28, 2026. In it, 34 percent say they ever use AI chatbots for at least one of eight health and medical reasons, and 15 percent of all U.S. adults use them to help decide whether to go to a doctor. Accessed October 25, 2026.
  2. A second survey. KFF, “Poll: 1 in 3 Adults Are Turning to AI Chatbots for Health Information, Equaling the Share Who Use Social Media for Health,” March 25, 2026, KFF Tracking Poll on Health Information and Trust, 1,343 U.S. adults, February 24 to March 2, 2026, in which 32 percent of adults turned to AI chatbots for health information in the past year. Accessed October 25, 2026.
  3. Clicks and summaries in Google search. Athena Chapekis and Anna Lieb, Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results,” July 22, 2025, browsing data from 900 U.S. adults, March 1 to 31, 2025, covering 68,879 Google searches. A click on a traditional result link came in 8 percent of visits with an AI summary and 15 percent of visits with traditional results only, and a click on a link inside the summary in 1 percent. Browsing ended on 26 percent of pages with an AI summary and 16 percent of pages with traditional results only. About one in five searches produced an AI summary, and 60 percent of searches that began with who, what, when or why did. Pew notes that the AI summary for any given search may change over time. Accessed October 25, 2026.
  4. How Google’s AI answers find and credit pages. Google Search Central, “AI features and your website,” last updated December 10, 2025. AI Overviews and AI Mode may run several related searches and link to supporting pages; the best practices for search remain relevant, and new machine-readable files, AI text files or markup are unneeded; the basics include allowing crawling in robots.txt and by any CDN or hosting, keeping important content in text and keeping Business Profile information up to date; robots.txt rules for Googlebot control crawling for Search; clicks from AI features count in the overall search traffic in Search Console, within the Web search type. Accessed October 25, 2026.
  5. OpenAI’s crawlers. OpenAI, “Overview of OpenAI Crawlers.” OAI-SearchBot surfaces websites in ChatGPT search results, and sites that opt out drop out of those answers, though they can still appear as navigational links; GPTBot crawls content that may be used in training; each setting is independent of the others. For ChatGPT-User, which acts when a user asks, “robots.txt rules may not apply.” Accessed October 25, 2026.
  6. Counting visits from ChatGPT. OpenAI Help Center, “Publishers and Developers – FAQ.” ChatGPT adds the parameter utm_source=chatgpt.com to referral URLs from its search results. Accessed October 25, 2026.
  7. New-patient flags in plan directories. Simon F. Haeder and Wendy Y. Xu, “Inaccuracies related to accepting new patient flags in Affordable Care Act plans’ provider directories,” Health Affairs Scholar 3(5), qxaf082, April 15, 2025, a secret shopper survey of Marketplace plans in six specialties and seven states, May 2022 to December 2023, which found that 14.5 percent of 20,514 verified listings wrongly said a provider was accepting new patients. Accessed October 25, 2026.
About jscTEK

jscTEK is a commercialization partner for life sciences and healthcare. We bring commercial strategy, brand and growth built on how healthcare buyers decide, backed by more than four decades of combined experience.

More insights

Keep reading.

  • A finance leader and a clinical leader review printed charts at a table in a bright office, his pen on one bar.

    Healthcare

    Margin before marketing

    What does each line keep from a new patient, before you spend to win one?

    Read the piece
  • A web page with product information and a carton and vial on the left, and safety information beside it on the right.

    Life sciences

    What an FDA-regulated product asks of a website

    What changes for your website the day your product is approved?

    Read the piece
  • A care coordinator in navy scrubs takes notes on a call at her desk, with a referral page open on her screen.

    Healthcare

    What referral sources need from a provider’s website

    What do the people who send you patients look for when they open your site?

    Read the piece

Start with the milestone.

A launch, a market entry, a platform that has to hold, or a case you have to make. Tell us what is on the horizon, and we can talk through the commercial, brand and digital work it calls for.