Signal of the week

A 10,000-person global survey documents a structural shift in when and how patients make healthcare decisions, and the exam room has already changed

The ZS Impact Institute published its inaugural 2026 Future of Health Report on June 24, 2026, drawing on responses from more than 10,000 healthcare consumers and licensed physicians across the United States, Germany, and China. The report documents a fundamental change in the patient-clinician dynamic: patients are making clinical decisions, including treatment preferences and medication choices, before they interact with a clinician.

The U.S. findings:

→ 58% of U.S. patients research symptoms before deciding whether to book a clinical appointment

→ 52% now arrive at clinical encounters requesting specific medications or therapies by name

→ 68% of healthcare providers report an increase in patients requesting specific therapies, confirming the patient-side data from the clinician's perspective

→ 90% of patients who use AI and digital health tools trust the information they receive nearly as much as advice from their own physician

The economic case for addressing this:

45-68% of patients actively delay entering the formal healthcare system until symptoms become too disruptive to ignore. ZS estimates earlier diagnosis across major disease areas could generate nearly $500 billion in annual direct medical savings in the United States, driven by intercepting conditions before acute or late-stage presentation. Jon Roffman, ZS: “Patients are changing faster than the system designed to serve them.”

The clinical evidence check

Mixed — AI health advice accuracy varies substantially by condition; patient trust is not calibrated to accuracy

How accurate is AI health advice compared to physician guidance?

A 2025 JAMA Internal Medicine study found that large language model responses to consumer health questions were rated by physicians as accurate or mostly accurate 72% of the time, and potentially harmful in 7% of cases. That 7% harmful response rate, applied to the hundreds of millions of consumer AI health queries generated daily, represents a meaningful patient safety signal.

AI performs well on factual information about common, well-documented conditions (hypertension, type 2 diabetes, common medications) and substantially worse on rare conditions, complex drug interactions in polypharmacy cases, and patient-specific risk stratification that requires EHR context the AI tool does not have access to.

The 90% trust finding from the ZS report is the most clinically significant data point: patients trust AI health information nearly as much as their own physician, regardless of whether the AI's accuracy for their specific clinical question is comparable to a physician's. Patient trust in AI health advice is not calibrated to AI's actual performance in the specific clinical context in which the patient is using it.

Regulatory radar

Practitioner voice

68% of healthcare providers report a measurable increase in patients requesting specific therapies by name, directly confirming what the patient survey found. The report also documents a persistent misalignment between clinicians and patients on medication adherence: patients cite confusion and cost anxiety as the primary drivers of non-initiation; physicians tend to attribute non-adherence to other factors.

This misalignment is not new, but AI health information is amplifying it. Patients who have researched treatment options arrive with cost expectations and side-effect concerns that were not part of the traditional pre-appointment conversation. Clinical teams that do not surface those concerns during the encounter will lose adherence they never knew they were at risk of losing.

Deal sheet

The Clinical Download  ·  clinicaldownload.com  ·  Manage preferences  ·  Unsubscribe