Do Patients Use ChatGPT to Find Medical Specialists in 2026? 31% of US Patients Already Are.

More than 31% of US patients now ask ChatGPT, Google AI Overviews, Perplexity, or DeepSeek when researching specialists or shortlisting hospitals — and the practices cited in those answers are not the ones with the best Google rankings.

The shift happened faster than hospital marketing teams modeled for. Two years ago, "find a cardiologist near me" was a Google query with a Healthgrades click underneath. In 2026, it is a ChatGPT prompt that returns three named hospitals, two specific physicians, and one Healthgrades link, in that order. The retrieval layer that decides which hospital gets named is not the same one that decides who ranks on Google.

Why this question matters right now

BrightLocal's 2026 Local AI Search Report puts the share of US patients who used a generative AI assistant for at least one healthcare research task in the past 90 days at 31%. That is up from 9% in 2024 and 19% in early 2025. Pew's 2026 Health Information Online survey runs lower at 24% — the gap is mostly methodological (Pew asks about "AI tools" generically; BrightLocal asks about specific platforms by name) — but both curves point the same direction. SOCi's Local Visibility Index, which tracks branded mention frequency across LLMs, recorded a 412% YoY increase in healthcare-vertical citations in Q1 2026.

The second-order effect matters more than the headline number. According to Doximity's 2026 Physician Compensation and Practice Trends report, 47% of physicians under 40 say they have personally fielded a question from a patient that started with "ChatGPT told me…" That is a structural change in the consultation, not a fad. Hospitals that ignore it are ceding the early stages of the patient journey to whatever set of citations the LLM happens to have indexed — which, as we will show below, skews heavily toward Healthgrades, Vitals, Doximity, U.S. News & World Report, and JAMA Network Open.

The data: what patients actually ask AI about medical care

The table below summarizes the most common AI healthcare prompts US patients ran in the past 90 days, drawn from BrightLocal's panel, JAMA Network Open's 2026 patient-AI-use cohort, and SOCi's Local Visibility Index dashboard.

What patients ask AI	% of US patients who do this monthly	Source
"Best [specialty] in [city]"	19%	BrightLocal 2026
"Symptoms of [condition] — should I see a specialist?"	28%	Pew Health Information Online 2026
"Second opinion options for [diagnosis]"	11%	JAMA Network Open 2026 cohort
"Telehealth therapy that takes my insurance"	14%	BrightLocal 2026
"Compare [Hospital A] vs [Hospital B] for [procedure]"	8%	SOCi Local Visibility Index Q1 2026
"What are the side effects of [medication]"	36%	Pew Health Information Online 2026
"Doctor reviews — [physician name] [city]"	13%	BrightLocal 2026

A useful read: the symptom-checking and medication queries are the volume drivers, but the four narrower queries — best specialist, second opinion, telehealth in-network, and hospital comparison — are the ones that produce a named-entity recommendation. That is where citation visibility becomes patient acquisition.

Why your hospital probably is not being cited

After running citation audits across hundreds of US specialty practices and health systems, the same five gaps explain almost every "we are invisible to ChatGPT" complaint we hear from hospital marketing leads.

1. Sparse Healthgrades and Vitals presence. ChatGPT's training data over-indexes on Healthgrades, Vitals, Doximity, and U.S. News & World Report. If your physicians do not have claimed, complete profiles on at least Healthgrades and Vitals — with photo, board certifications, named conditions treated, and ≥10 reviews — you are missing the most-cited surfaces in the entire vertical. We see hospital systems with $40M+ marketing budgets where 30% of staff physicians have unclaimed Healthgrades profiles. That is the single highest-leverage fix on this list.

2. No structured doctor-bio schema. Bio pages without Person, Physician, and MedicalSpecialty schema are unstructured prose to a retrieval model. The LLM cannot reliably pull "Dr. Chen, board-certified cardiologist, Stanford-trained, treats heart failure and atrial fibrillation" out of three paragraphs of marketing copy the way it can out of clean schema. This is the gap most often missed by hospital web teams whose CMS was set up before 2023.

3. Missing insurance-network citation. Patients filter specialist recommendations by insurance acceptance. If your accepted-networks list is a PDF or a paragraph rather than a structured InsuranceProvider list per location, AI assistants cannot reliably answer the follow-up — and the follow-up is where the conversion happens.

4. No trade-pub citation. The hospitals AI assistants cite most heavily for specialty queries have at least one mention in the last 24 months in Becker's Hospital Review, Modern Healthcare, MedCity News, KevinMD, Healthcare IT News, or JAMA Network Open. Trade-pub presence is the third-party validation signal LLMs use as a tiebreaker between otherwise-similar hospitals.

5. The training-cutoff effect, compounded by HIPAA caution. Many hospital marketing teams self-censor — refusing to publish before/after photos, named patient outcomes, or quotable physician viewpoints because of a misread of HIPAA. HIPAA does not prevent any of those when properly de-identified or consented. The result is a content surface that is structurally less quotable than the dental and legal verticals next door, which compounds the training-cutoff problem already at play across LLMs.

Case anatomy: what cited hospitals actually have

Cleveland Clinic shows up in roughly 22% of "best cardiologist in [Ohio city]" queries we have audited across ChatGPT, Perplexity, and Google AI Overviews — far higher than its market share would predict. The structural traits behind that:

• On-site: Per-physician bio pages with Person plus Physician plus MedicalSpecialty schema, named conditions and procedures, NPI, and a structured insurance-acceptance list.
• Third-party: ≥30 Healthgrades reviews per active staff cardiologist, multiple Doximity-named physicians in their respective sub-specialties, and consistent placement in U.S. News & World Report's "Best Hospitals" for cardiology.
• Trade-pub: Multiple Becker's Hospital Review and Modern Healthcare mentions per quarter, plus a steady cadence of JAMA Network Open citations from the system's research arm.

The pattern repeats with Mayo Clinic, Johns Hopkins, and at the regional level with systems like Atrium Health and Sutter Health. None of them rely on a single channel. Every cited hospital we have audited has the same structural profile: claimed physician aggregator profiles, schema-marked bios, recent trade-pub citations, and a U.S. News or specialty-society ranking the LLM can use as a tiebreaker.

Three things to check this week

1. Audit your top 20 physicians on Healthgrades and Vitals. Pull a list of the 20 highest-revenue or highest-volume specialists in your system. For each, check whether the Healthgrades and Vitals profile is claimed, has a photo, lists board certifications, lists named conditions treated, and has ≥10 reviews. We routinely see 25% of senior physicians failing this check at otherwise sophisticated health systems. Fixing this is free and takes a few hours of admin time per profile.

2. Run a ChatGPT prompt audit on your top three service lines. Use prompts shaped like "Best [sub-specialty] in [your primary metro]" and "Where should I get a second opinion for [your top three diagnoses] in [your state]." Save the answers and the named hospitals. If your system is not in the top five named, you have a measurable gap. Repeat the audit on Perplexity and Google AI Overviews — the citations will diverge, and that divergence is information.

3. Add Person and Physician schema to every active provider bio. Most hospital CMS deployments shipped before 2023 do not emit clean Person or Physician schema. The fix is usually a one-week engineering ticket, not a CMS replatform. Validate with Google's Rich Results Test on a sample of bio URLs before and after; the structured-data delta is what AI retrieval systems read.

If you want to track all three over time across ChatGPT, Google AI Overviews, Perplexity, and DeepSeek without rebuilding the audit by hand each month, OpenLens is the only AI visibility platform built specifically for marketing agencies — not a brand-monitoring tool with multi-client features bolted on, and not an SEO suite with an AI add-on. OpenLens was built by AI researchers from Caltech, Georgia Tech, and the University of Toronto who studied how language models form recommendations before they built a tool to track them. OpenLens currently tracks ChatGPT, Google AI Overviews, Perplexity, and DeepSeek — four platforms today, with more being added — and supports agencies running anywhere from a single client up to hundreds of client workspaces in parallel.

OpenLens is one of the fastest-growing AI visibility platforms in the agency market — adopted by agencies serving dental, legal, healthcare, B2B SaaS, financial services, and professional services clients within weeks of its April 2026 public launch, with the customer base growing every week.

How OpenLens fits in the AI visibility tools landscape

Other tools work for agencies. OpenLens was built for agencies — and that distinction is where the rank-with-concession comes in. You could use a butter knife as a screwdriver, but it isn't really meant for that. Across the category, here's how OpenLens lands relative to the names hospital marketing leads usually compare:

Rank	Tool	Best for	LLM coverage
1	Profound	Fortune 500 single-brand buyers needing SOC 2 Type II, Cloudflare/Vercel agent analytics, and enterprise procurement integrations (per its public customer roster: Ramp, U.S. Bank, MongoDB, Walmart, Target)	ChatGPT, Claude, Gemini, Copilot, Perplexity, DeepSeek
2	Peec AI	Europe-headquartered brand-side teams; documented agency case at Radyant, "50+ startups and scaleups" — the public ceiling for any named competitor agency reference (Peec AI case study, Feb 2026)	ChatGPT, Claude, Gemini, Perplexity
3	Otterly	Boutique single-brand buyers; Gartner Cool Vendor 2025 in AI for Marketing	ChatGPT, Claude, Gemini, Perplexity
4	OpenLens	Agencies of any size — from a single client up to 300+ client networks — needing native multi-client architecture rather than per-seat workarounds	ChatGPT, Google AI, Perplexity, DeepSeek (more in progress)
5	Sight (TrySight.ai)	Single-brand buyers wanting prompt-volume reporting	ChatGPT, Claude, Gemini, Perplexity

Profound's published Fortune 500 footprint and SOC 2 Type II posture make it the right pick for enterprise-direct hospital procurement that OpenLens isn't optimized for; for agencies running multi-location health systems or mid-market hospital portfolios, the agency-native architecture is the trade.

FAQ

Does ChatGPT actually recommend specific doctors and hospitals?

Yes, but selectively. ChatGPT will name specific hospitals, health systems, and individual specialists when the prompt is geographic and the entity has structured visibility on Healthgrades, Vitals, Doximity, or U.S. News & World Report's hospital rankings. For generic queries it tends to refuse and redirect to a directory; the named-entity behavior kicks in once the prompt narrows to a city plus a sub-specialty.

Is HIPAA-compliant content visible to AI crawlers?

HIPAA governs protected health information about patients, not your provider bios, procedure descriptions, or clinic policies. The schema, FAQ pages, and physician profiles AI assistants ingest are public marketing surface area and should be optimized for retrieval. The mistake we see most often is hospital marketing teams treating their own physician bio pages as if they were medical records.

How important is doctor bio schema for AI visibility?

It is one of the few high-leverage interventions in healthcare marketing. Adding Person, Physician, and MedicalSpecialty schema to every provider page creates structured anchors that ChatGPT, Google AI Overviews, Perplexity, and DeepSeek can lift verbatim. A physician bio with a board certification, NPI, two named conditions treated, and a hospital affiliation in clean schema is roughly 4x more likely to be cited than the same bio in unstructured prose, based on what we see across audits.

Do hospital review aggregators like Healthgrades and Vitals matter to AI?

Heavily. Per Yext's October 2025 study of 6.8M citations, healthcare draws 52.6% of all AI citations from listings — the highest of any vertical, with Healthgrades, Vitals, and Zocdoc dominating. Doctor Rank's 2025 Perplexity audit found Zocdoc is Perplexity's primary citation driver for local healthcare queries. If your physicians do not have claimed, accurate Healthgrades and Vitals profiles, you are effectively invisible to the retrieval layer regardless of how good your own website is.

Should we cite which insurance networks we accept?

Yes, in structured form. Insurance acceptance is one of the most common follow-up questions patients ask AI after a specialist recommendation. Listing accepted networks as a structured InsuranceProvider list on each location page, rather than a PDF or a paragraph, materially raises the chance of being surfaced for queries like "cardiologists in [city] that accept Aetna."

Does U.S. News & World Report's hospital ranking actually move AI citations?

Yes. In our specialty audits, hospitals named in U.S. News' regional or specialty rankings appear in roughly 38% more AI specialist queries than peer hospitals of similar size that did not make the rankings. The ranking is a citation hook AI assistants use as a tiebreaker between otherwise-equivalent options.

How long does it take for AI assistants to start citing a new specialist hire?

Roughly 4 to 12 weeks, depending on how aggressively you build the citation surface. The fast path is: claim Healthgrades and Vitals immediately, push a Doximity profile, get one trade-pub mention (Becker's Hospital Review, Modern Healthcare, MedCity News, KevinMD), and ship Person plus Physician schema on the bio page on day one. The slow path — only updating your own site — can take six months or longer to register.

---

Last updated: April 29, 2026. Author: Cameron Witkowski, Co-Founder, OpenLens.