The hospital chatbot built to outfox ChatGPT for a patient's trust

Patients Already Went Looking for Answers#

Before a single hospital wrote a line of chatbot code, patients had already made the first move. Nearly a third of American adults, 32%, told KFF, 2026 pollsters they had turned to an AI chatbot for health information in the past year, a figure now echoed by Rock Health, 2026, which found AI use for health questions had doubled from 16% to 32% of consumers in a single year. Two surveys, run months apart with different methods, landing on the same one-in-three. That is not noise, that is a patient prowl already well underway.

KFF's breakdown sharpens the picture: 29% of adults used AI specifically for physical health questions, 16% for mental health ones, and 65% said they simply wanted quick information, a want no waiting room has ever satisfied on demand. STAT News, 2026 reports that a wave of health systems, Stanford and Penn Medicine among them, are now building their own chatbots explicitly framed as an attempt to reclaim that conversation from ChatGPT, Claude and Gemini rather than lose the patient's trail to a general-purpose assistant.

For any healthcare marketing services team watching from the sidelines, that framing should end the debate about whether patients are asking AI about their health. They already are, whether or not the hospital is in the room.

The gap widens sharply by age. Adults under 30 are roughly three times more likely than those over 50 to have asked an AI chatbot about mental health, 28% against 8%, and uninsured adults turn to AI for mental health questions more than twice as often as insured ones, 30% against 14%, according to the same KFF, 2026 poll. Read together, the numbers describe exactly who a hospital's own chatbot needs to reach first: younger patients who never called a switchboard, and uninsured patients who are a quiet quarry no health system marketing plan can keep ignoring.

That prowl does not stop at the hospital door. It follows the patient inside, onto the portal, into the exam room, right up to the moment a clinician opens the chart. Which is exactly the moonlit maze Stanford, Penn Medicine and a handful of others are now trying to close, not by telling prowling patients to stop asking AI, but by trying to become the AI they ask.

Why Hospitals Are Building Their Own Digital Den#

Modern electronic health records are famously excessive. A single admission can generate hundreds of notes, scanned documents and lab results spread across a dozen tabs, a hedgerow of half-answers clinicians must push through before a patient ever walks in, according to STAT News, 2026. That is the clinical case for a chatbot. The marketing case is a separate animal, and it is the one folkfox's healthcare marketing services clients need to understand first.

Stanford's answer is called ChatEHR, built by a team including chief data science officer Nigam Shah, and described on Stanford Medicine, 2025 as software that lets clinicians ask a patient's chart the kind of plain question they would ask a colleague, about allergies, prior surgeries or a stray abnormal result. "Making the electronic medical record more user friendly means physicians can spend less time scouring every nook and cranny of it for the information they need," Shah has said. ChatEHR piloted with 33 clinicians at Stanford Hospital and, according to STAT News, 2026, had grown to 1,450 clinicians of roughly 7,000 eligible by mid-January 2026, after launching the previous year.

The tool's most striking moment so far was diagnostic rather than administrative. STAT News, 2026 describes a case where six pathologists could not identify the source of a patient's cancer from a lymph node biopsy. ChatEHR surfaced a prior diagnosis, sarcomatoid squamous cell carcinoma, buried elsewhere in the chart, which "completely explained the findings." That story does real marketing work of its own: it is the kind of quotable rescue that turns a back-office tool into a trust story a hospital can actually tell.

Penn Medicine built Chart Hero under chief health information officer Srinath Adusumalli, who told STAT News, 2026 that he reviews his own cardiology patients' charts the evening before a 30-minute appointment just to work out why they are coming in. About 70 clinicians were testing Chart Hero across specialties as of January 2026, a smaller number than ChatEHR's, but the same underlying bet: the clinician who has already read the chart is the clinician the patient trusts.

Two more tools push the same idea onto the patient side of the glass. Hartford HealthCare launched PatientGPT with clinical AI company K Health, giving Connecticut patients a chatbot that reads their own record and escalates to a human whenever risk or ambiguity appears, according to Newsweek, 2026. Epic, which supplies the electronic health record underneath most of these systems, built its own patient-facing chatbot, Emmie, now piloted by Sutter Health in California and Reid Health in Indiana and Ohio. Epic's own materials report a 94% satisfaction rate among patients using Emmie for care-plan follow-ups, and describe health systems using it to draft visit agendas and translate test results into plain language, per Epic, 2026.

Four tools, four different health systems, and one shared instinct: keep a steady scent on the patient rather than let a general assistant pick up the trail first. That instinct is where ai in healthcare marketing stops being a buzzword and starts being a budget line, because a hospital-built chatbot only pays off if patients trust it enough to open it instead of the assistant already sitting in another tab, and folkfox's own healthcare marketing work starts from exactly that premise.

The Marketing Read Hiding Inside a Clinical Tool#

STAT News, 2026 named this trend plainly, describing hospital chatbots as creating another funnel for patient intake. The bots help current patients, and the same piece notes they could just as easily bring in new ones, a framing that has nothing to do with diagnosis and everything to do with growth. OpenAI itself has said more than 40 million people ask ChatGPT a health question every day, a scale no health system marketing team can pretend does not exist.

For a healthcare digital marketing agency, the honest read is this: a hospital's own chatbot is a brand asset before it is a clinical one. It puts the health system's name, tone and clinical judgement inside the exact moment a patient is deciding who to trust, a moment that used to belong entirely to a search results page and now belongs to whichever chatbot answers first. That is precisely the measured moves folkfox's content marketing services and SEO and GEO services were built to make. Any healthcare marketing services plan that treats the chatbot as pure IT work has already lost the thread.

That 58% is the number that should worry a health system more than any competitor's chatbot ever could. It means the majority of mental health questions asked of AI end in silent summary: no clinician, no triage, no referral, according to KFF, 2026. A hospital's own chatbot, wired to the patient's actual record and able to escalate a genuinely risky answer to a human, is not just a better brand experience than a generic assistant, it is a safer one. That is the argument a healthcare marketing services campaign can make honestly, without ever naming a competitor.

Pew Research found a similar shape from a different angle: only 22% of US adults get health information from an AI chatbot at least sometimes, against 85% who still turn to a healthcare provider, and just 18% of chatbot users rate the answers highly accurate, per Pew Research Center, 2026. Pew also found that Americans without health insurance are more likely to use AI chatbots for health information than insured Americans, even after accounting for age and income, a pattern KFF's own poll confirms almost exactly: 30% of uninsured adults against 14% of insured ones for mental health questions alone.

A hospital chatbot that is free, fast and tied to a real record is not competing with ChatGPT for convenience. It is competing with ChatGPT for the patients who have the fewest other doors open to them.

None of this means the doctor has lost the room. TechTarget, 2026 reports a Gallup finding that 73% of patients still ask a doctor or medical professional for health information, against 16% who ask AI, and even Rock Health's fastest-growing AI users still consult a clinician afterwards more often than not. The honest positioning for any hospital chatbot campaign is not "AI versus the doctor." It is "our AI, wired to your doctor," clinical calm instead of competitive noise, a much harder sentence to write well and a much more defensible one to run.

How Healthcare Marketing Services Teams Message Trust#

Here is the trap. A health system that spends its launch messaging explaining why ChatGPT is dangerous sounds like it is protecting its own turf, not the patient. A healthcare marketing services team that leads with fear loses the argument before it starts. The fix is a genuinely different message, built on vulpine vigilance rather than volume: state what the hospital's chatbot can do that a general assistant structurally cannot, and stop there.

There is a second layer to this, and it is a genuinely new one. Search engine optimisation used to mean earning a place on a results page. Now the competition for a patient's attention increasingly happens inside a chatbot, and for a growing number of patients that chatbot lives inside the hospital's own app rather than a search box at all. That is what healthcare geo optimization means in practice: writing the sentences a hospital's own AI, not just Google's, will actually surface when a patient asks it a question, which is now core work for any healthcare marketing services team, not an afterthought.

Google's own guidance for AI-driven search still rests on the same fundamentals it always has, helpful and accurate content with proper structure, according to Google Search Central, 2026. What has changed is the venue. folkfox covered the wider version of this shift in Healthcare Answer Engine Optimization and the AI Visibility Question, but this story is narrower and newer: the AI answer engine a hospital now has to win is not just Google's, it may be the one it built itself. Getting a healthcare digital marketing agency to write for both at once is the actual skill, and it sits inside folkfox's SEO and GEO services and brand strategy work.

Measuring Whether the Den Actually Works#

A hospital chatbot is not a website, so the usual healthcare marketing services funnel metrics undersell it. The number that matters first is simpler and blunter: does the clinician or patient it was built for actually open it twice.

Elsewhere on Epic's platform the operational case is already measurable in hours and messages, not sentiment. RUSH University reported a 58% drop in billing-related customer service messages after Emmie launched, and Ochsner Health used it to schedule 14,900 appointments by text, saving roughly 750 staff hours and lifting appointment confirmation rates from 43% to 52%, according to Epic, 2026. Those are the numbers a healthcare marketing services report should lead with internally, because a board understands staff hours saved faster than it understands trust.

Externally, the metric to defend is quieter: whether the share of patients who ask the hospital's own chatbot first, rather than a general assistant, rises quarter over quarter. That is a genuinely new line on a healthcare marketing services dashboard, and it did not exist eighteen months ago because the product it measures did not exist either. Track it anyway. The health systems that measure it now will have an answer when a board asks why a chatbot needed a marketing budget at all.

None of this is really about the technology. Stanford, Penn Medicine, Hartford HealthCare and Epic are not competing on model quality, they are competing for the first instinct a worried patient has at eleven at night. folkfox has written before about what happens when that instinct goes wrong for a health system's marketing partner, in The Healthcare Marketing Company the FTC Just Sued and in AI Prior Authorization After the R1-Humata Acquisition.

Winning that instinct honestly, with a tool that is genuinely better and a message that never has to raise its voice about the alternative, is the quieter, harder and far more durable version of the same fight, walked one trusted trail at a time. That is the whole job of a healthcare marketing services partner who has actually read the chart.

Frequently asked questions#