Healthcare AI spending is approaching $37 billion globally in 2026 and 85% of healthcare organizations say they plan to increase their AI budgets this year. The bulk of that investment is concentrated in workflows that happen during and after the clinical encounter: ambient documentation, clinical decision support, medical coding, prior authorization and revenue cycle management. NVIDIA’s 2026 State of AI in Healthcare survey found that the top use cases across providers and payers are administrative tasks and workflow optimization and Menlo Ventures reported the largest share of provider AI spending going to ambient documentation and coding. These use cases mainly target efficiency and administrative burden, not clinical outcomes. Some genuinely streamline the system and add value, while others simply modify payments rather than improve care. What none of them touch is the decision that matters most to the member: choosing the right provider for their specific care needs.

Meanwhile, provider-level quality measurement has matured considerably. Validated methodologies can now score individual physician performance against evidence-based guidelines across complete multi-payer claims data. The clinically validated measurement has been developed. What has been missing is a way to put it in front of the member at the moment they need it most: when they are sick and trying to determine what kind of care they need, whom they should see and which provider is best equipped to help them.

Healthcare AI’s blind spot

Choosing a provider is the key decision that sets the trajectory for a member’s entire care experience: the diagnosis, the medical care, the recommended procedures, the referrals to other doctors and the associated costs that accumulate. Choices made by providers influence roughly 80% of the healthcare dollar, making which provider a member chooses a critical decision. Yet, choosing which physician to see for a new condition remains largely untouched by the AI investment now flowing into healthcare.

Consumers know this moment matters. PwC’s 2025 US Healthcare Consumer Insights Survey found that 53% of consumers are already using or interested in AI-powered care navigation tools that recommend the right provider or care setting. Among Gen X and Millennials, that figure reaches 73%. Zocdoc’s 2025 data shows the average patient compares 21 provider profiles before booking an appointment. They read reviews, check credentials, filter by availability and insurance acceptance. In Zocdoc’s Censuswide survey, patients ranked personal connection and online reviews as top factors in their provider selection. Clinical performance did not make the list.

AI is already influencing these choices. Rater8’s 2025 research found that 26% of patients say AI tools directly influenced their choice of provider, nearly matching the influence of primary care referrals at 28% and healthcare review sites at 29%. The demand for AI-guided provider selection is clear. The problem is that the AI consumers are using today has no clinical quality data behind it.

Variation at the provider level

What many members don’t know is physician quality and adherence to evidence-based guidelines vary significantly from provider to provider and that variation shapes both their outcomes and their costs. For example, two orthopedic surgeons evaluating the same knee pain in comparable members may recommend entirely different courses of action. One follows conservative, evidence-based guidelines including physical therapy and non-surgical options first. The other proceeds immediately to surgery, even when it hasn’t been shown to help. Across more than 230 million members, Embold Health’s measurement methodology has found that surgical rates for the same condition can differ by more than a factor of 30 between the conservative, evidence-based providers and providers practicing more aggressive, non-evidence-based medicine. A single unwarranted operation can cost tens of thousands of dollars and expose the member to complications, recovery time and the need for future procedures, all from an intervention that should never have occurred.

The tools members currently rely on to choose between those two surgeons (Google results, online reviews, star ratings and wait times) do not reflect any of this. A member might reasonably assume the physician with a three-month wait is the best one in their area, when in reality, some of the highest-quality physicians have next-week availability simply because they haven’t built a wide-reaching reputation yet. Reputation and clinical quality measure different things. That difference shows up in varying outcomes, downstream utilization, avoidable complications, unnecessary procedures and the total cost of care.

AI at the point of decision

Quality measurement alone, no matter how rigorous, cannot change outcomes if it never reaches the member at the point of choosing a provider. For years, clinical quality data has served retrospective purposes: public reporting, value-based purchasing and readmission penalties. It evaluated care that was already delivered. The member choosing a provider had no access to it and the systems they relied on (health plan directories, search engines and review platforms) were built around geography, contract status and convenience. The data existed in one place and the decision happened in another.

AI closes that distance. A traditional provider search takes “my knee hurts” and returns a list sorted by proximity or alphabetic order. An AI-powered experience can take that same input, ask clarifying questions about symptoms, history and severity, apply clinical context to determine the right type of provider the member needs. It can also weigh the member’s preferences, such as gender, location, system of care and virtual or in person. Then it connects the member to a physician whose measured performance shows they excel in that specific area of medicine. The clinical quality data does the analytical work and AI delivers it through a personalized, conversational experience that helps members choose with confidence.

McKinsey’s 2025 Consumer Health Insights Survey found that consumers who use AI-enabled healthcare tools, including tools to get matched with a provider, report satisfaction rates of 54% compared to 30% among non-users. When the right data is behind the AI, the experience changes. The highest-value application of AI in healthcare is at the start of the care journey,  where AI, paired with the right clinical data, does the work of sorting through complexity so the member doesn’t have to. That empowers members to choose the right provider with confidence.

What the industry should demand

Members are already using AI to select providers and that adoption will accelerate. The question is whether the AI guiding those decisions will be built on clinical evidence or on the same inputs that have always been available: proximity, reputation, volume and reviews.

For AI at the provider selection moment to produce different outcomes, it requires a different foundation: clinically validated provider quality data, risk-adjusted for patient complexity and evaluating appropriateness alongside outcomes. And taking the time to clinically validate the AI tool itself is critical to its success.

Individual provider level quality measurement has matured and AI is the tool that finally puts it to work at the moment it can change outcomes, before the appointment is booked, before the referral is made and before the cost is incurred. The investment should match the opportunity.