OpenEvidence Is Becoming the HCP Search Engine. Here’s What Pharma Should Do About It.

For years, the HCP digital playbook assumed a clinician with a question would search Google, land on a brand’s HCP site, and maybe download a resource. A study published in JAMA on October 15 shows how much of that behavior now happens somewhere else.

Clinician AI, by the numbers

In 2.5 years, OpenEvidence became where clinicians ask what to do next

952,008

US clinicians submitted queries, including 556,200 physicians

271M

queries from January 2024 through June 2026

703,265

monthly active users in June 2026, up from 355

33.1%

of physician queries were about treatment and management, vs 13.9% diagnosis

What the study found

Researchers from Harvard Medical School and Brigham and Women’s Hospital analyzed NPI-verified US clinicians on OpenEvidence from January 2024 through June 2026 (JAMA, Oct 15, 2026). Over those 2.5 years:

• 952,008 clinicians, including 556,200 physicians, submitted 271 million queries.
• Monthly active users grew from 355 to 703,265.
• Queries per active user per month more than doubled, from 18.0 to 39.5.

What clinicians ask matters more than how many are asking. Among physician queries, 51.9% were about clinical decisions, and treatment and management made up 33.1%, compared with 13.9% for diagnosis. The authors note that nurse practitioners and physician assistants asked for much the same things.

Clinicians aren’t mainly using OpenEvidence to work out what a patient has. They’re using it to decide what to do next. That’s the moment where your indication, line of therapy, dosing, and safety information either show up accurately or don’t.

What the study doesn’t tell us

It’s worth reading this study with the same skepticism you’d bring to any vendor data. Three co-authors are OpenEvidence employees, and the company supplied the de-identified data and contributed to the analysis, though it didn’t fund the work. The JAMA Network also licenses its journal content to OpenEvidence, which the accompanying editorial discloses.

More important for pharma: the study didn’t evaluate whether the answers were accurate, and it couldn’t separate point-of-care use from chart prep or self-education. The authors say so directly: how well these tools handle treatment selection and medication guidance, the tasks that dominate real use, remains largely unmeasured. We know clinicians are asking. Nobody has measured what they’re being told. That gap is exactly where brand teams should focus.

Who is asking

The heaviest users aren’t the HCPs most brand plans are built around.

• Nurse practitioners and physician assistants each averaged 44.5 queries per active month, ahead of physicians at 37.1.
• Physicians within a decade of graduation used it at 2.54 times their share of the workforce, and residents and fellows at 1.84 times. Physicians 20 or more years out came in at 0.31.
• Oncology and hematology (2.00x), internal medicine (1.88x), and family medicine (1.80x) led among specialties. Procedural and diagnostic specialties trailed: ophthalmology 0.23x, radiology and orthopedics 0.24x, anesthesiology 0.37x.

Who is asking

Early-career physicians and generalists ask AI the most

Share of OpenEvidence physician queries relative to share of the US physician workforce. Above 1.0x means a group asks more than its size alone would predict.

By career stage

• Within 10 years of graduation 2.54x
• Residents and fellows 1.84x
• 20+ years from graduation 0.31x

By specialty

• Oncology and hematology 2.00x
• Internal medicine 1.88x
• Family medicine 1.80x
• Anesthesiology 0.37x
• Orthopedics 0.24x
• Radiology 0.24x
• Ophthalmology 0.23x

Above 1.0x Below 1.0x Dashed line: 1.0x = proportional to workforce

Why the usual HCP playbook doesn’t carry over

OpenEvidence builds its answers from licensed medical literature and society guidelines, with in-line citations. Brand websites aren’t in that source set. The pattern is spreading: on October 20, Doximity and Wolters Kluwer announced that UpToDate content will appear inside Doximity Ask, rolling out over the coming weeks to clinicians at health systems that use both.

That changes what drives visibility. Schema and on-page SEO on a brand’s HCP site still matter for Google and for general-purpose AI tools like ChatGPT. They won’t move an answer in OpenEvidence. What moves those answers is the evidence base itself: the label, the pivotal publications, the guidelines, and how clearly each states what your therapy is for.

In practice, that puts medical affairs and publications planning at the center of HCP findability. Brand and SEO teams can’t own this alone, and in most cases shouldn’t.

Where the answer comes from

The clinician’s question no longer passes through your HCP site

The old path

Every step shows up in your analytics

1. Clinical question
   What do I prescribe next?
2. ↓
3. Google search
   Paid and organic results
4. ↓
5. Brand HCP site
   Indication, dosing, safety
6. ↓
7. Resource download
   Tracked as a conversion

The new path

None of it shows up in your analytics

1. Clinical question
   Often outside their specialty
2. ↓
3. Clinician AI tool
   OpenEvidence, Doximity Ask
4. ↓
5. Synthesized answer
   Built from journals, guidelines, the label
6. ↓
7. Treatment decision
   No brand site visit required

Visibility now depends on the evidence base, not the brand site.

Advertising, and the compliance lane

OpenEvidence is free to clinicians because it sells targeted ads to pharma and device companies. That makes it a real media channel, with real targeting potential given what the study shows about who asks what. According to New York Times reporting, ads appear on roughly 5% of queries, shown in the few seconds while the answer generates. There’s still little independent data on how those placements perform. Treat it as a channel to test, not a substitute for being represented accurately in the answer itself. And an ad there is promotional like any other, so fair balance and OPDP rules apply in full.

The harder question is what brand teams should do when an AI answer gets something wrong. Two rules keep you on the right side of the line:

• Monitoring is fine. Influencing is not. Commercial teams can and should know what these tools say about their brand. Trying to shape clinician-facing answers, especially on unapproved uses, crosses into territory that belongs to medical affairs.
• Corrections go through scientific channels. Inaccurate or off-label answers should be routed to medical affairs and handled through publications, medical information, and scientific exchange. FDA’s guidance on how firms can address misinformation about their products is the relevant reference point.

Get this wrong and a visibility project becomes a regulatory problem. Get it right and medical affairs becomes your strongest ally in the AI search era.

The measurement blind spot

None of this activity shows up in your analytics. A clinician who gets a complete treatment answer in OpenEvidence often never clicks through to your HCP site. HCP site traffic was already a weak proxy for HCP interest. It’s getting weaker every month.

If your dashboards treat HCP sessions as the primary demand signal, they’re undercounting how clinicians actually learn about your therapy. Two fixes, one quick and one strategic.

The quick fix: make AI referrals visible. The small share of clinicians who do click through from AI tools usually land in GA4 as generic referral traffic. Create a custom channel group for AI referrers so they’re reported on their own:

Channel: AI Assistants
Condition: Session source matches regex
.*(openevidence|doximity|chatgpt|openai|perplexity|gemini|copilot|claude)\..*

Place it above the default Referral channel in the group order, or those sessions will still be absorbed into Referral.

The strategic fix: measure the answer, not the click. Track clinician AI visibility as its own metric set:

What pharma brands should do now

1. Run a clinician AI audit. Ask OpenEvidence, Doximity Ask, and ChatGPT the treatment questions HCPs actually ask in your category. Cover your brand and your competitors.
2. Check every answer against the label. Look at indication, line of therapy, dosing, and safety. Route anything outdated or off-label to medical affairs rather than trying to fix it from the brand side.
3. Start a baseline now. Save the questions and answers, then re-run them on a set schedule. These tools change constantly, and you can’t reconstruct what they said last quarter.
4. Prioritize by who’s asking. Weight the audit and your targeting toward NPs, PAs, early-career physicians, and the high-usage specialties in your category.
5. Bring medical affairs in early. The levers that move these answers sit in publications and scientific exchange, not in brand creative.
6. Measure it separately. Report clinician AI visibility as its own metric set, not something inferred from HCP site traffic.

The bottom line

Clinicians are increasingly getting treatment answers before they reach a brand site, and nobody has yet measured whether those answers are right. The brands that come out ahead will know what the answers say about them and will make sure the evidence behind them is accurate and easy to find.

PharmaForward runs clinician AI audits for a small number of biopharma teams each quarter. If you want to see what OpenEvidence says about your brand today, let’s talk.

Frequently asked questions

What is OpenEvidence?
An AI clinical information platform that answers clinicians’ free-text questions with citations to peer-reviewed journals and society guidelines. It is free to verified clinicians and supported by pharma and device advertising.

How many clinicians use OpenEvidence?
A JAMA study published October 15, 2026 found 952,008 US clinicians submitted 271 million queries between January 2024 and June 2026, with monthly active users reaching 703,265.

Does schema markup on a brand’s HCP site affect OpenEvidence answers?
No. OpenEvidence draws on licensed medical literature and guidelines, not brand websites. Schema still helps with Google and general-purpose AI tools.

Can pharma companies advertise on OpenEvidence?
Yes. OpenEvidence sells targeted ads to pharma and device companies. Those ads are promotional and subject to the same fair balance and OPDP requirements as any other channel.

What should a brand team do if an AI tool gives an inaccurate or off-label answer?
Document it and route it to medical affairs. Corrections should go through scientific channels such as publications and medical information, not brand-side efforts to influence the answer.

How can pharma brands measure visibility in clinician AI tools?
Run a recurring audit of category treatment questions and track share of answer, on-label accuracy, cited sources, and change over time. In GA4, a custom AI Assistants channel group captures the clinicians who do click through.

Sources

• Patel et al. Clinician Use of an Artificial Intelligence–Enabled Clinical Information Platform. JAMA, Oct 15, 2026
• JAMA Editor’s Note: Engaging With Medical Evidence in the Age of AI
• Harvard Medical School: First Large-Scale Study Shows How Doctors Use AI in Medicine
• Becker’s Hospital Review: OpenEvidence, 6 things to know (citing The New York Times, June 2026)
• Wolters Kluwer: Doximity and UpToDate partnership, Oct 20, 2026