The 2026 State of AI Visibility
How ChatGPT, Claude, Gemini, Google AI Overview, and Perplexity answer patient-intent questions — why the household-name national brands almost never appear, how regional systems dominate the top of every leaderboard, and why "best hospital" queries return a completely different answer than "best cardiologist in Portland."
Executive summary
Hospitals are unlike any other category we've measured. There is no single national winner. Instead, AI segments the answer by geography and by specialty — and the difference between the two rankings is stark.
The 60-second read
Unlike SaaS or DTC categories where one brand dominates, healthcare has no single hospital or system cited as the default answer. The top-ranked hospital in our dataset — Maine Medical Center — pulls only 8.5% of all mentions. Mayo Clinic, the household-name national brand most people expect to lead here, sits at #16 with less than 1% share. Regional systems dominate every top slot. This is a fundamentally different competitive shape than any commercial category.
Add a city, state, or region to any query and the leaderboard flips. "Best cardiac surgeon in Dallas" surfaces UT Southwestern, Baylor, and Medical City Dallas. "Best hospital in Maine" surfaces Maine Medical Center and Northern Light Mercy. AI segments by geography before it segments by anything else — ~98% of the top-cited hospitals in the dataset are regional systems, not national brands. In 5,156 answers analyzed across ChatGPT, Claude, Gemini, Google AI Overview, and Perplexity, the "national default" is a myth for hospital queries.
Hospitals sub-segment by geography before they sub-segment by anything else — and even within a metro, the leaderboard fragments further by specialty. A Dallas cardiac query surfaces UT Southwestern and Baylor; a Dallas pediatric query surfaces Children's Medical Center Dallas (99% positive sentiment, one of the cleanest brand profiles in the entire dataset). These are effectively different categories with different competitive landscapes — measured together, they hide the truth.
AI models cite medicare.gov (Hospital Compare) and healthgrades.com more than any hospital's own domain when answering patient-intent queries. US News rankings and Newsweek's Best Hospitals list are the next tier. If your health system doesn't appear well on Hospital Compare, you're invisible upstream of every AI answer.
Health queries are YMYL (Your Money Your Life). AI models weight sources with clinician-reviewer bylines, medical credentials, and government citations disproportionately when answering. Health-system content without a "Medically reviewed by [Dr. X, MD]" byline gets demoted regardless of how much SEO investment sits behind it.
Where ChatGPT and Claude will directly name a "best" hospital, Google AI Overview is conservative — it defers to symptom-checkers, CMS data, and MedlinePlus more often than to specific institutions. When Google AI Overview does name a hospital, it heavily prefers those with public quality metrics on file at CMS.
Methodology
We ran a full patient-intent benchmark against five AI surfaces — ChatGPT (gpt-5.6), Claude Sonnet 5, Gemini 3.6 Flash, Google AI Overview, and Perplexity Sonar. Queries were sourced from real patient search intent: “best hospital for cardiac care in [region],” “where should I go for cancer treatment,” “when to switch pediatricians,” “hospital near me with the best maternity ward,” “specific vs Kaiser Permanente for outcomes.” That’s 5,156 completed AI answers, parsed by an LLM extractor for: which hospitals and systems were mentioned, in what position within ranked answers, with what sentiment framing, and citing which source URLs.
Queries were split across the intent shapes patients actually use:
Mentions were tagged for sentiment, position within ranked answers, and the specific URLs the AI referenced. Everything below is derived from that dataset.
Scan period: 2026-07-26 to 2026-08-26. Category defined by Foxish's benchmark taxonomy: "hospital." Report generated August 27, 2026. Next update: September 30, 2026.
The national leaderboard
This is what an aggregate leaderboard hides more than it reveals. Read it, then read the specialty and regional breakdowns further down — they tell a much sharper story about who's actually winning what.
| Rank | Hospital / System | Share of voice | Mentions | Sentiment |
|---|---|---|---|---|
| 1 | Maine Medical Center | 1,448 | 91% positive | |
| 2 | Magnolia Regional Health Center | 1,291 | 86% positive | |
| 3 | UT Southwestern Medical Center | 1,126 | 94% positive | |
| 4 | Northern Light Mercy Hospital | 938 | 82% positive | |
| 5 | North Mississippi Medical Center | 685 | 80% positive | |
| 6 | Baylor University Medical Center | 610 | 96% positive | |
| 7 | Baylor Scott & White Health | 559 | 91% positive | |
| 8 | MaineHealth | 489 | 78% positive | |
| 9 | Children’s Medical Center Dallas | 386 | 99% positive | |
| 10 | Medical City Dallas | 361 | 93% positive |
The long tail: 793 hospitals show up in 1–3 queries each. These are almost always regional systems responding to location-modified queries. See the geographic breakdown for what those look like.
Where the answer flips
Three regional clusters — Dallas, Maine, and Mississippi — carry 53% of all mentions in the dataset. Nothing else even approaches those shares. Once a query has a location grip, AI answers cluster tightly around the systems those markets already recognize.
The Dallas cluster is the single largest regional share. On any Dallas-modified query, the top-5 is drawn from the same six systems — UT Southwestern, Baylor UMC, Baylor Scott & White, Medical City Dallas, Children’s Medical Center Dallas, and Texas Health Presbyterian. Sentiment on Children’s Medical Center Dallas is 99% positive — the cleanest brand profile of any hospital we measured.
Maine Medical Center is the #1 hospital overall — not because it’s a national brand, but because Maine queries are heavily represented in the benchmark set and MMC and Northern Light Mercy own that geography. MaineHealth (the parent system) and Barbara Bush Children’s Hospital fill out the top-6 for the state.
Magnolia Regional Health Center (MRHC) and North Mississippi Medical Center are essentially the two names Mississippi queries return. On MS-modified prompts, they alternate top-2 across all five AI sources — a two-hospital duopoly for the state’s AI answer set.
The most striking finding. Mayo Clinic — the household-name hospital most people expect to lead any "best hospital" query — sits at #16 with 130 mentions. Cleveland Clinic and Johns Hopkins don’t appear in the top 30 at all. This isn’t how commercial categories work; it’s not even how SEO works. It’s AI following patient-intent language, which is almost always local.
Hospitals are the first category we’ve measured where the aggregate leaderboard is actively misleading. Mayo Clinic and Cleveland Clinic look like defaults if you’re thinking about buyer perception — they’re essentially invisible in the AI answer set. The real competitive landscape is 50+ regional sub-categories, each with its own 3–5 hospital shortlist.
Per-model breakdown
The specialty leaders are largely consistent — but the confidence with which each model names them varies dramatically. Google AI Overview is the most conservative; ChatGPT the most willing to give a specific answer.
Google AI Overview is a fundamentally different surface for health queries. It hedges toward government sources (CMS, MedlinePlus, NIH) and away from naming a specific institution. If you're trying to appear in AI answers to health questions, ChatGPT and Claude are the surfaces to prioritize — Google AI Overview is going to defer to public quality data first.
The citation graph
The hospital’s own domain is the single biggest driver. Five of the top six most-cited domains are hospital .org sites; US News, Healthgrades, and CMS round out the top 15 as the ranking-site layer. If your service-line pages aren’t comprehensive, no amount of PR closes the gap.
Health content is treated by AI as Your Money Your Life — a category where AI models actively down-weight sources without medical credentials, government citations, or clinician-reviewer bylines. A hospital site with rich SEO but no "Medically reviewed by [Dr. X, MD]" bylines gets demoted. This is a fundamentally different signal set than any commercial category.
US News, Healthgrades, and CMS (medicare.gov) together carry ~17% of top-25 domain citations — the ranking-site layer. But the bigger driver is your own .org: five of the top six most-cited domains belong to individual health systems. Getting comprehensive on your own site is table stakes; getting listed favorably on US News and Healthgrades is the accelerator.
Reading the tea leaves
Add specifics. "Best hospital in the US" gets you Mayo Clinic. "Best cardiologist in Portland who accepts BCBS" gets you a shortlist of three specific programs you'd never see otherwise. AI is doing sophisticated segmentation but only if the query gives it the grip to work from. And check the sources it cites — if it's leaning on CMS Hospital Compare, that's a good sign; if it's leaning on Reddit threads, treat the answer skeptically.
Your dominance is specialty-anchored. Cleveland Clinic wins cardiac; MSK wins oncology; HSS wins orthopedic. The moat is your specialty. Your risk is any query with a location modifier — you're often not the answer, even in cities where you have facilities. The buyer needs to add "or Cleveland Clinic in [city]" to your name, and AI often won't do that work for them.
You aren't competing with Cleveland Clinic on national queries. You're competing with UT Southwestern, Medical City Dallas, Baylor, and the other regional systems in your market — for location-modified queries and for insurance-specific queries. The plays here are (1) get comprehensive on CMS Hospital Compare, (2) publish specialty-comparison content that names other regional systems by name, and (3) make sure your service-line pages have clinician-reviewer bylines.
The playbook
Adapted for the hospital / health system category specifically. Ordered by cost and effort.
Health content is YMYL. AI models demote content without visible clinician review. Every "Cardiac care at [Hospital]" page should have a "Medically reviewed by [Dr. Name, MD]" byline plus ClaimReview or MedicalWebPage schema. This is the single highest-impact fix for health-content visibility — and most systems haven't done it.
Expected lift: substantial across specialty queries. Cost: 1-2 weeks of editorial + template work.
CMS Hospital Compare is the top AI citation source for health queries. Missing quality measures, low participation in patient-experience surveys, or gaps in condition-specific data hurt AI visibility upstream of anything you do on-site. Fill every measure your system is eligible for. This is one meeting away from a project plan.
Expected lift: foundation for all downstream AI visibility. Cost: internal coordination, no vendor spend.
The location-query market rewards head-to-head content. A "UT Southwestern vs Baylor Scott & White for cardiac care" page (published by either party) targets the exact query shape patients ask. AI models pull comparison-format pages heavily for regional queries. Do this respectfully and factually — inaccurate content gets discounted by AI models.
Expected lift: closes location-modified query gaps in your metro. Cost: 1-2 weeks per comparison page.
Healthgrades and US News are consistently in the top-5 AI citation sources. Both allow you to submit and update quality data, physician rosters, and specialty programs. Most health systems have stale or partial data on both. The fastest way to get cited more is to make the sources AI relies on reflect your actual capabilities.
Expected lift: 5-15 additional citations per month across queries in your region. Cost: 2-4 weeks of coordinated data collection.
Every fix takes 2-6 weeks to show up in AI citations. Without weekly tracking on the specific queries you care about — "best hospital in [your metro]," "best cardiologist in [your metro]," "in-network [specialty] for [major payer]" — you're guessing about causation. Fixing CMS data doesn't move national queries. It moves regional ones. You need the resolution to see it.
Expected lift: attribution clarity. Cost: 15 minutes a week.
About this report
Foxish is an AI visibility monitoring platform. We track how ChatGPT, Claude, Gemini, Perplexity, and Google AI Overview answer patient-intent questions across health-system categories — brand mentions, sentiment, share of voice, and the citation graph behind every answer. This report is a snapshot of what Foxish sees for the hospital / health system category. If you'd like the same view for your specific metro or specialty — cardiac, oncology, orthopedic, or the specific location you serve — the underlying platform runs your custom prompt set on a schedule and generates a live version of every chart in this report.
Data collected via Foxish's benchmark pipeline: batch API calls to OpenAI, Anthropic, Google (Gemini), plus scraped Google AI Overview responses. Category defined by Foxish's benchmark taxonomy: "hospital." Report generated August 27, 2026. Next update: September 30, 2026.
Get your region
The aggregate national leaderboard is directionally interesting but the metro or specialty you're actually competing in is what matters. Send us your metro and top 3-5 competitor systems — we'll run Foxish's benchmark pipeline against your regional prompt set for 7 days and generate a report in this exact shape with your system's position included.
Or try Foxish yourself at foxish.ai/signup — the same pipeline that produced this report, watching your health system's visibility on a schedule you set.