Which top clinical decision support startups win doctors?

Are doctors actually using AI clinical decision support every day?

AI clinical decision support is already part of everyday medicine for a growing number of doctors, particularly when they need to check medical evidence, make sense of patient information, or document a consultation.

Doximity's 2026 State of AI in Medicine survey, covering 3,151 U.S. physicians, found that 54% were using AI in clinical practice. Across its spring 2025 and winter 2025–2026 survey waves, reported adoption increased from 47% to 63%. These were different survey samples, so the increase measures a broader trend rather than changes in the behavior of exactly the same doctors.

The details are more revealing. AI-assisted medical literature searches rose from 22% to 35% of surveyed physicians, while ambient documentation increased from 20% to 29%. Doctors were using AI to find medical information more frequently than to write clinical notes, despite the enormous attention given to AI scribes.

OpenEvidence offers a more direct measure. The company reported approximately 20 million clinical consultations with its AI in January 2026 and announced one million consultations in a single day that March. Those figures suggest substantial repeat usage, although the one-day record doesn't tell us the platform's normal daily average.

What doctors are adopting also varies considerably. Some deliberately open an AI tool to investigate a diagnosis. Others encounter AI suggestions inside their hospital's electronic health records. That distinction becomes important when comparing which startups doctors actually prefer.

What does it mean for a clinical decision support startup to win doctors?

Winning doctors means getting them to use a clinical AI product repeatedly and trust its answers, although hospital-wide deployments can produce substantial usage without doctors choosing the software themselves.

OpenEvidence and Glass Health compete directly for physicians looking for clinical answers. Regard and Abridge work inside hospital systems, where administrators often choose the software before individual doctors encounter it. Aidoc analyzes imaging workflows and alerts medical teams, sometimes without requiring a physician to open a separate AI application.

A doctor voluntarily returning to OpenEvidence several times a week tells us something different from 5,000 hospital employees having access to Abridge. Both are commercially valuable, but only the first directly demonstrates individual product preference.

We therefore give the greatest weight to recurring physician activity, actual acceptance of clinical suggestions, and evidence that usage continues after initial deployment. Hospital customer counts and contract announcements help establish reach, but they cannot tell us how much doctors like the product.

Has OpenEvidence already won the race for individual doctors?

OpenEvidence is currently the clearest leader among independent clinical AI startups when it comes to physicians actively choosing and repeatedly using a decision-support tool.

Its growth has been extraordinary. According to data compiled by Sacra, monthly clinical consultations increased from approximately 3 million in early 2025 to 20 million in January 2026, a roughly 6.7-fold increase. Its subsequent record of one million consultations in a single day suggests that usage continued to climb.

OpenEvidence also reports that more than 65% of U.S. physicians use its platform monthly and over half use it daily. Those numbers come from the company itself rather than an independent measurement of the entire physician population. Still, the disclosed consultation volumes support the picture of a product with unusually strong repeat engagement.

Its commercial growth is nearly as striking. Sacra estimated approximately $300 million in annualized revenue by July 2026, compared with $150 million at the end of 2025. A September financing reportedly valued OpenEvidence at $15 billion.

The business model helps explain the adoption. Verified physicians can use the product for free, while OpenEvidence makes money primarily from pharmaceutical and medical-device advertising. Doctors don't need to persuade a hospital procurement department before trying it.

For individual physician adoption, OpenEvidence has the strongest publicly available case. Its biggest unanswered question concerns clinical outcomes: widespread usage and trust are impressive, but neither establishes that its recommendations consistently lead to better treatment decisions.

If you want more recent data on this point, please see our latest clinical decision support market report.

Can Glass Health seriously challenge OpenEvidence?

Glass Health has a compelling AI product for clinical reasoning, but OpenEvidence remains far ahead on publicly disclosed physician usage.

Glass helps doctors work through symptoms, possible diagnoses, guidelines, and treatment plans. A physician can describe a complicated patient case and ask the software to explore what conditions deserve attention. OpenEvidence, meanwhile, has built much of its popularity around quickly retrieving and explaining relevant medical evidence.

There is considerable overlap, particularly as medical AI tools become better at both literature retrieval and clinical reasoning.

Glass has published results from clinical evaluations, including the Harvard-Stanford ARISE NOHARM benchmark, which examines the safety of recommendations in medical scenarios. Such tests help establish what a model can do under controlled conditions. They say much less about how often doctors return to the product after trying it.

OpenEvidence discloses millions of monthly consultations, while Glass hasn't published a similarly detailed series covering active physicians, consultation frequency, and repeat usage.

Glass could still appeal to doctors who want more help developing a differential diagnosis rather than finding an answer to a specific clinical question. For now, however, there is no convincing public evidence that it comes close to OpenEvidence in physician adoption.

Is Abridge becoming more than an AI note-taking tool for doctors?

Abridge is turning its huge hospital distribution network into a serious advantage in clinical decision support, although its newer AI capabilities still have less proven engagement than its documentation product.

By October 2025, Abridge reported adoption by more than 200 health systems. In August 2026, the company said more than 300 enterprise health systems had adopted its context-aware clinical decision-support capabilities following their April introduction. The announcements have slightly different scopes, so they shouldn't be treated as an exact like-for-like growth calculation.

Individual hospital networks show what that reach means. UPMC announced plans to expand Abridge to more than 12,000 clinicians by 2026. WVU Medicine separately planned deployment to more than 2,800 clinicians across 25 hospitals and associated care locations.

These are rollout targets and institutional adoption figures, not verified daily active user counts. Even so, Abridge has secured something many medical AI startups struggle to obtain: direct access to thousands of doctors inside major healthcare systems.

Abridge's newer clinical intelligence features connect patient conversations, medical records, and relevant evidence. Instead of typing a separate question into an AI search engine, clinicians can receive information related to the patient they are already treating.

That makes Abridge particularly dangerous to standalone clinical AI products. Doctors may prefer receiving the right information inside their existing workflow, provided the recommendations are accurate and useful.

If you want more recent data on this point, please see our latest clinical decision support market report.

Is Regard actually helping doctors catch more diagnoses?

Regard has some of the strongest evidence that physicians are accepting AI-generated diagnostic suggestions inside hospitals, although its clearest measurable benefits concern documentation and revenue capture.

Regard examines electronic health records, identifies potentially relevant conditions, and prepares diagnoses and documentation for physician review.

By March 2026, the company reported deployments across more than 150 hospitals and over 12 million clinician-accepted diagnoses. It also claimed to have helped customer health systems unlock more than $200 million in earned revenue through improved diagnosis documentation.

These are cumulative company-reported figures. An accepted diagnosis may reflect better documentation of a condition the physician already recognized, so 12 million accepted diagnoses shouldn't be interpreted as 12 million previously missed diseases discovered by AI.

The physician adoption numbers at Monument Health are particularly interesting. According to a presentation involving the health system at the 2026 American Hospital Association Leadership Summit, adoption increased from the high 60% range to 90% after Regard introduced its Max agent.

Monument also reported a 52% improvement in hierarchical condition category capture and a 44% improvement in complication and comorbidity capture. Those measures are important for documenting patient complexity and supporting reimbursement, though they do not directly establish better patient outcomes.

Regard has since expanded beyond hospital medicine into cardiology and surgery. Its software clearly fits physician workflows and improves clinical documentation. Whether it also produces major reductions in missed diagnoses or treatment errors remains a separate question.

Is Aidoc winning hospitals without asking doctors to change their habits?

Aidoc has built one of the largest clinical AI footprints by analyzing medical imaging and bringing findings directly into hospital workflows, often without requiring doctors to open a separate application.

The company reports use of its AI technology at more than 1,600 hospitals worldwide. Its platform helps identify relevant imaging findings, prioritize cases, alert clinicians, and coordinate follow-up care.

A radiologist reviewing scans may encounter Aidoc-generated information automatically, rather than deciding whether to consult an AI assistant. That gives Aidoc an advantage over products that must persuade busy physicians to initiate a separate interaction.

In August 2026, Aidoc announced a diagnostic AI consortium involving 12 U.S. health systems, including Cedars-Sinai, Houston Methodist, Northwell Health, and Sutter Health. Together, the participating organizations serve nearly 20 million patients annually. That is the combined reach of the health systems, not the number of patients whose examinations Aidoc actually analyzed.

Aidoc's distribution is impressive, but we cannot compare 1,600 hospitals directly with OpenEvidence's millions of consultations. One measures institutional reach; the other measures individual interactions.

Aidoc already has a strong position in automated clinical support. Its challenge is demonstrating how consistently its alerts change physician decisions and improve care, particularly when measured across different hospitals and specialties.

Which clinical decision support startups have the most doctors using them?

OpenEvidence has the strongest disclosed evidence of active physician-led consultations, while Abridge, Regard, and Aidoc lead on different measures of hospital adoption and clinical workflow activity.

We've compared the operating figures companies make public, while keeping their limitations visible. Some report actual interactions, others count hospitals, and several rely on deployment targets or selected customer studies. There is no credible way to turn these different measurements into one universal number of active doctors.

Company	Published adoption evidence	What the figures actually show
OpenEvidence	About 20M monthly consultations in January 2026; 1M on a peak day in March	Very large volume of physician-initiated clinical questions
Abridge	More than 300 health systems adopting contextual clinical decision support	Extensive enterprise distribution, with active usage of newer features less clear
Aidoc	More than 1,600 hospitals using its technology	Broad deployment of automated clinical AI
Regard	More than 150 hospitals; 12M+ accepted diagnoses	Repeated physician interaction with AI-generated diagnostic documentation
Ambience Healthcare	92% clinician adoption in MultiCare's evaluation; 80% utilization at Houston Methodist	Strong adoption across selected enterprise deployments
Pathway / Doximity	300,000+ quarterly AI-tool users reported by Doximity	Large existing physician audience, now inside a public company
VisualDx	More than 2,300 hospitals and institutions reported	Broad reach for established visual diagnostic support
Atropos Health	More than 10,000 clinical questions converted into studies	Meaningful usage for specialized evidence generation
Glass Health	Clinical evaluation results, but no comparable active-user series disclosed	Product capability with less transparent adoption scale

Do doctors actually prefer Ambience Healthcare over competing AI tools?

Ambience Healthcare has produced unusually convincing evidence of clinician preference in head-to-head hospital evaluations, particularly for documentation and related clinical workflows.

MultiCare Health System compared three ambient AI products during 2025, involving more than 550 clinicians across over 20 specialties. When the health system announced its decision in February 2026, it reported that Ambience achieved 92% clinician adoption and a net promoter score 63 points higher than the next-best product.

This is more informative than a company announcing that another hospital purchased its software. Clinicians used competing products during the same purchasing process, giving the hospital a basis for comparing acceptance.

Other deployments tell a similar story. Ardent Health reported 90% clinician utilization during a pilot covering 17 specialties and seven languages. Houston Methodist reported 80% utilization across specialties and a 40% reduction in documentation time.

We can't average those percentages into an industry-wide adoption rate because the organizations used different evaluation methods and clinician populations. But the consistency across several substantial hospital deployments makes the results harder to dismiss as an isolated success.

Ambience also introduced commercial arrangements linked to measurable outcomes, tying its value more closely to clinical and financial improvements.

The limitation is the product category. Most published evidence concerns documentation, coding, and integrated workflows. Ambience has clearly persuaded doctors to use its software, but its position in clinical decision-making specifically is less established than its position in ambient documentation.

If you want more recent data on this point, please see our latest clinical decision support market report.

Could Doximity and Pathway take doctors away from OpenEvidence?

Doximity could become OpenEvidence's most serious competitor for physician attention because it already reaches a huge professional audience and acquired the clinical reference startup Pathway.

Doximity purchased Pathway in 2025 for a reported $63 million. Pathway had developed structured clinical information covering guidelines, medications, and important medical studies, and previously reported hundreds of thousands of healthcare professionals using its tools.

The acquisition placed those capabilities inside Doximity's broader physician platform. During the October–December 2025 quarter, Doximity reported more than 300,000 unique clinicians using its AI tools, which it described as approximately 20% of U.S. doctors. That number includes its wider AI offering, not exclusively clinical question-answering.

Doximity also published a 2026 comparison based on 1,315 side-by-side physician evaluations. It reported that DoxGPT won 61% of preferences when physicians selected a better answer, compared with 26% for the closest competitor. Because Doximity ran and published the analysis, we should treat it as company-sponsored evidence rather than an independent ranking of medical AI systems.

OpenEvidence currently has the more impressive disclosed volume of clinical consultations. But Doximity has years of physician relationships and can distribute AI tools alongside services doctors already use.

Pathway's acquisition also raises an interesting question for smaller startups: joining an established physician network could be a faster way to reach doctors than trying to build a comparable audience independently.

Are OpenEvidence and Glass Health replacing UpToDate and VisualDx?

OpenEvidence and other AI-native clinical tools are winning physician attention, but there is not yet enough evidence that doctors or hospitals are abandoning established references such as UpToDate and VisualDx.

VisualDx reports deployment across more than 2,300 hospitals and institutions worldwide, including over half of U.S. academic medical centers. Its diagnostic tools combine medical knowledge with a substantial visual reference library, particularly useful for dermatology, infectious diseases, and other conditions where appearances help distinguish diagnoses.

UpToDate has even stronger institutional economics. Sacra estimated approximately $690 million in 2025 revenue, with around 80% coming from institutional customers. Hospitals have spent years integrating established references into their procurement processes and clinical workflows.

OpenEvidence's rise creates a different kind of competition. Doctors can ask a question, receive a synthesized answer, and inspect linked medical evidence without paying a personal subscription. That convenience can gradually reduce the number of questions directed to traditional references.

But a doctor could still use VisualDx to examine a difficult rash, UpToDate for a comprehensive clinical review, and OpenEvidence for a quick question about recent treatment guidance.

The battle will become clearer when we can measure changes in paid subscriptions, renewals, and usage across competing tools. For now, AI-native startups have clearly captured physician activity, while the established providers retain major institutional relationships.

If you want more recent data on this point, please see our latest clinical decision support market report.

Can Atropos Health answer medical questions that ChatGPT and OpenEvidence cannot?

Atropos Health stands out because it can produce new clinical evidence from real patient data when existing medical literature doesn't answer a doctor's specific question.

Its Green Button service takes clinical questions and runs retrospective studies using real-world healthcare datasets. Doctors can investigate outcomes in patient populations with particular diseases, treatments, or combinations of medical conditions.

Atropos reports more than 10,000 clinical questions converted into studies and over 50 peer-reviewed publications connected to its work. Its traditional Green Button service targets results within 48 hours, which makes it better suited to complex evidence requests than quick questions during a consultation.

Consider a physician treating a patient with several chronic conditions who doesn't closely resemble the participants in a major clinical trial. OpenEvidence may help retrieve and interpret published studies, while Atropos can potentially investigate outcomes among more comparable patients in available datasets.

That capability has clear value, though the scale is quite different. Ten thousand completed studies cannot be compared directly with OpenEvidence's millions of monthly consultations, and the companies serve different kinds of questions.

Atropos looks particularly well placed for specialized clinical research and difficult treatment decisions. It has less evidence of becoming an everyday tool for large numbers of individual physicians.

Are clinical AI startups giving doctors accurate answers or just convenient ones?

The best clinical AI tools have demonstrated strong performance in physician evaluations, but accuracy, source quality, and safety still vary enough that convenience alone cannot explain which products doctors should trust.

A June 2026 research preprint examined 620 real point-of-care questions submitted to OpenEvidence across 30 specialties, alongside 187 additional benchmark questions. Researchers recruited 149 practicing physicians from 36 U.S. states to compare responses from OpenEvidence and several general-purpose AI systems.

OpenEvidence performed best across five evaluation dimensions: accuracy, usefulness, source quality, verifiability, and completeness. In the primary analysis, the difference between its win and loss rates ranged from 25 to 39 percentage points.

That's meaningful comparative evidence, especially because practicing physicians judged questions resembling their normal clinical work. However, the original question sample came from OpenEvidence, and the study doesn't establish how every competing specialist medical AI product would perform.

Other evaluations focus on dangerous mistakes rather than the overall usefulness of answers. The Stanford-Harvard ARISE NOHARM benchmark tests how clinical AI responds to cases where inappropriate advice could harm patients. Glass and Doximity have both highlighted results from versions of their systems, though comparisons depend heavily on the model versions and evaluation methods.

Clinical reliability remains a major unresolved competitive advantage. Doctors may initially adopt a tool because it responds quickly, but sustained reliance will depend on whether it provides dependable evidence, handles uncertainty appropriately, and avoids dangerous recommendations.

Have clinical decision support startups actually improved patient outcomes?

Clinical AI startups have strong evidence of saving doctors time and improving documentation, but considerably less evidence that their products consistently reduce medical errors or improve patient health.

Regard reports millions of clinician-accepted diagnoses and stronger documentation of patient conditions. Ambience has demonstrated shorter documentation times and substantial clinician adoption. Abridge has secured broad institutional rollout. These are important achievements, although they don't directly measure whether patients recover faster or experience fewer complications.

A September 2026 research preprint provides a useful illustration of what stronger clinical evidence could look like. Researchers studied an intervention providing drug-interaction information across approximately 2.81 million prescriptions and 1,700 physicians in India. The intervention was associated with an 8.6% reduction in drug-drug interaction errors.

The research wasn't a test of OpenEvidence, Regard, or Abridge. It did, however, show that giving physicians relevant information at the right moment can change prescribing behavior, even when the alerts are optional.

For the major startups, the most convincing future evidence would come from prospective comparisons showing fewer missed diagnoses, medication errors, avoidable hospitalizations, or other measurable patient harms.

Several clinical AI startups have clearly changed doctors' workflows. Claims about better patient outcomes still need more direct evidence.

Can OpenEvidence, Regard, and Abridge make serious money from clinical AI?

Clinical decision support is already supporting substantial businesses, but OpenEvidence, Regard, and Abridge make money in very different ways.

OpenEvidence primarily sells advertising tied to physician engagement. Sacra estimated that its annualized revenue reached approximately $300 million in July 2026, about twice its estimated level at the end of 2025. These are third-party estimates rather than audited financial statements.

Regard makes its case to hospitals through better clinical documentation and financial returns. The company's reported $200 million in additional earned revenue for customers indicates the size of the economic benefits it claims to deliver. That figure is customer revenue unlocked, not Regard's own sales.

Abridge and Ambience sell into health systems that want to reduce administrative work and improve clinical workflows. Aidoc earns money through enterprise deployment of imaging AI and care-coordination software. Atropos charges for specialized evidence services, while Glass Health offers free and paid clinical reasoning capabilities.

Looking across these companies, physician usage and customer spending often happen in different places. A doctor may freely choose OpenEvidence, while a hospital independently pays for Regard or Abridge because the financial and operational benefits justify the contract.

Company	Main business model	What customers pay for
OpenEvidence	Advertising and commercial partnerships	Access to physicians during clinical information searches
Regard	Enterprise hospital contracts	Diagnosis documentation, workflow improvement, revenue capture
Abridge	Enterprise health-system contracts	Documentation and contextual clinical intelligence
Ambience Healthcare	Enterprise contracts, including outcome-linked arrangements	Time savings, clinical documentation, coding improvements
Aidoc	Enterprise clinical AI deployments	Imaging analysis, triage, care coordination
Atropos Health	Enterprise evidence services and data products	Clinical research using real-world patient data
Glass Health	Free and paid offerings	AI-assisted diagnosis and clinical reasoning
VisualDx	Individual and institutional subscriptions	Diagnostic reference and visual medical knowledge

Could Epic and Microsoft take over clinical decision support from startups?

Epic, Microsoft, and other major healthcare software providers could take a large share of clinical AI usage because their products already sit inside the systems doctors use every day.

Electronic health records are particularly important. If useful clinical guidance appears directly in Epic or Oracle Health, physicians have less reason to open a separate application or manually transfer patient details.

Abridge has already moved in this direction by bringing patient-specific intelligence into existing EHR workflows. Regard has also pursued integration with Microsoft's Dragon Copilot environment, connecting patient-record analysis with clinical documentation.

OpenEvidence is approaching the same market from the opposite direction. It has built a large direct physician audience and can develop integrations from that position, rather than depending entirely on hospital procurement to gain access to users.

The risk is greatest for smaller companies offering capabilities that larger platforms can reproduce easily. A generic medical chatbot may struggle if an EHR vendor includes similar functionality in an existing subscription. Specialized tools with proven clinical performance, trusted evidence, or deeply integrated workflows should be harder to replace.

Clinical decision support will probably remain spread across several products, especially because imaging triage, diagnostic reasoning, documentation, and medical literature searches solve different problems. But controlling the physician's main workflow could become an enormous distribution advantage.

Which clinical decision support startups are actually winning doctors today?

OpenEvidence is the strongest winner in direct physician adoption, while Abridge, Regard, and Aidoc lead in different parts of hospital-based clinical decision support.

OpenEvidence has the clearest evidence of doctors independently choosing an AI clinical reference tool and returning frequently. Its consultation volume, reported physician reach, and rapid estimated revenue growth put it well ahead of smaller standalone challengers such as Glass Health.

Among enterprise startups, Abridge has extensive health-system distribution and is moving deeper into patient-specific clinical intelligence. Regard has demonstrated large-scale physician acceptance of diagnostic suggestions, with particularly strong evidence around documentation and coding. Aidoc has reached more than 1,600 hospitals through imaging analysis and automated clinical workflows.

Ambience Healthcare deserves recognition for high clinician adoption in competitive hospital evaluations, although its best-documented results concern documentation. Atropos serves a smaller but potentially valuable market for generating clinical evidence from patient data. Doximity, now incorporating Pathway, is a major competitive threat to OpenEvidence, even though it is no longer a startup.

Our comparison points to two distinct kinds of winners. OpenEvidence is becoming a default destination for doctors who want fast clinical answers. Abridge, Regard, and Aidoc are building clinical intelligence into the systems where doctors already work.

The most important unanswered question concerns patient outcomes. Physician adoption has moved much faster than independently verified evidence of fewer diagnostic mistakes, safer prescribing, or better treatment results. The startups that can demonstrate those improvements at scale will have a much stronger claim to lasting clinical leadership.

If you want more recent data on this point, please see our latest clinical decision support market report.