Physicians used to completely rely on their own clinical knowledge, consulting reference textbooks, attending CME conferences and searching the medical literature to answer difficult questions. Today, artificial intelligence tools can summarize research, generate board-style questions and help physicians navigate vast amounts of medical information. “I think [AI is] a very useful asset especially during clinic consultations where it can summarize encounters and dictate letters allowing more interaction time with the patient,” a general surgeon on Sermo notes.
Since AI is relatively new, many practicing physicians are learning as they go. “There was no such thing as AI when I was in medical school,” a pediatrician writes on Sermo. Doctors are learning how to balance and incorporate the tools without relying on them too heavily.
Many of those physicians would like future generations to have earlier preparation. In a small-sample poll, 94% of physicians on Sermo agreed that medical schools need dedicated AI courses. “Curriculum on AI bias, ethics, and standards needs to be an active part of medical education and clinical standard-setting,” an internist on Sermo shares.
If you’re already a practicing physician, going back to medical school to learn about AI tools likely isn’t the most practical route. You can still develop a practical understanding of how to fold AI tools into the work you already do through your own research, or if you want structured AI medical training, you can consider taking a class like Harvard Medical School’s dedicated AI in Clinical Medicine CME course. This article presents a guide to the AI tools physicians are using, how to judge whether a tool is reliable and how you can use AI for board prep and recertification.
The AI tools physicians are using for clinical knowledge
Physicians use AI tools ranging from premium platforms to free chatbots. In a poll, Sermo members said the tasks AI is most ready to support include personalizing treatment plans (19%) and supporting decision making at point-of-care (17%). A general practitioner member appreciates AI scribe tools in particular: “Using AI scribe has been life changing as a family physician,” they write. “Allows the focus to shift back to our patients.”
To start making use of AI’s capabilities now, these are tools you can consider adding to your workflow:
UpToDate Expert AI
Launched in late 2025, UpToDate Expert AI has been adopted by 70% of UpToDate’s largest enterprise systems according to the company. It delivers responses grounded exclusively in UpToDate’s content, and it now can award CME credits directly to physicians who engage in dialogue with the tool. In March 2026, UpToDate partnered with Microsoft to integrate into Dragon Copilot, Microsoft 365 Copilot and Microsoft Teams.
ChatGPT for Clinicians
OpenAI launched ChatGPT for Clinicians in April 2026, and it’s free for verified physicians, nurse practitioners, physician assistants or pharmacists. That raises a real question for physicians of whether a free tool could be accurate enough. A 2026 JMIR Medical Education study found ChatGPT had “moderate but meaningful alignment” with UpToDate, with the most similarities in pharmacology. The study concluded that ChatGPT for physicians is not a substitute for evidence-based resources but may serve as an “accessible adjunctive tool.”
Other AI tools for physicians
Several other AI clinical decision support tools are noteworthy. OpenEvidence, an official partner of publications like The New England Journal of Medicine, JAMA andCochrane Systematic Reviews, is a tool that provides evidence-based clinical answers and it’s free for verified health care practitioners.
Dyna AI is a generative AI capability for the clinical decision support tools DynaMedex and Dynamic Health. Specialty-specific tools also exist, like nucleai.ai for pathologists and Rad AI for radiologists.
How to evaluate whether an AI tool is reliable for clinical use
Not all AI tools are created equal. Generally, you can use these four criteria to evaluate any AI tool before trusting it at the point of care:
Source grounding
Ideally, tools cite primary sources, so you know where information came from. When “black box” AI tools don’t offer explanations for how they arrived at a conclusion, it’s harder for physicians to identify incorrect results, researchers argue. Some AI clinical decision support tools are grounded in peer-reviewed literature or expert-curated content (like UpToDate). Others use less rigorous means, generating responses from a general-purpose language model trained on the open internet.
“It would be foolish to ignore [AI], almost as much as it would have been foolish to reject ‘indirect’ auscultation—the stethoscope—just because it was newer than placing one’s ear on the chest,” an internist states on Sermo. “That noted, reliance on a ‘black box’ is dangerous, foolish, and legally, professionally, and morally undefendable.”
Accuracy and hallucination risk
AI tools can produce confident-sounding answers that are simply wrong—a problem known as “hallucination.” Many physicians are already aware of this; In a Sermo poll, the ability to easily edit, verify and override AI suggestions was the most common (23%) feature members said would help them feel comfortable using the tools.
A 2024 study found ChatGPT’s average medical reference accuracy was “weak/unrelated” compared to UpToDate in managing otorhinolaryngology and head and neck surgery clinical scenarios. This is why guides to responsible AI use typically recommend verifying AI-generated clinical recommendations against established evidence-based resources before applying them to patient care. “AI should not be presented as a panacea for all ills or as a tool to which we can delegate our critical thinking,” urges a pediatric neurologist on Sermo.
CME accreditation
Consider whether use of the platform qualifies for CME credit through an accredited CME program. Many general-purpose AI chatbots do not. If CME credit matters for your Maintenance of Certification (MOC) requirements, this distinction can save you a separate step.
How to use AI for board prep and recertification
AI can genuinely help with board prep and MOC if you use it the right way. Here’s a reliable approach for AI board prep:
Use AI for question generation and concept review, not as a primary study source
AI excels at generating practice questions, explaining concepts in different ways and identifying knowledge gaps. Still, it’s not a replacement for board review courses. Think of it as a study partner, not the syllabus.
Verify every clinical fact against an evidence-based source
Treat AI output as a first draft, not a final answer. Cross-reference clinical recommendations, drug dosages or guideline references against specialty guidelines or primary literature before you commit them to memory. “Clinicians need to both respect [AI’s] value and be on alert for its limitations,” an anesthesiologist on Sermo cautions.
Use AI to stay current between board cycles
AI tools that synthesize recent literature (such as OpenEvidence and UpToDate Expert AI) can help you track guideline changes, new evidence and emerging therapies in the years between formal board review periods.
Track CME credit through AI-integrated platforms
UpToDate Expert AI now awards CME credit within the AI workflow itself. When choosing among AI continuing medical education tools, look for ones that integrate CME tracking directly rather than forcing you to log credits separately.
How to integrate AI learning into your workflow
AI tools are designed to turn “staying current” from a separate task into something that happens as a byproduct of clinical work. “Medicine needs to start viewing integration with computer systems differently—not as two separate clinical and technical disciplines, but as something parallel, seamless, and appropriate,” argues a family medicine doctor on Sermo.
Sermo poll respondents are most interested in using AI to reduce documentation burden (26%) or summarize complex patient histories before visits (17%). “I believe artificial intelligence has significant potential to streamline medical documentation by enhancing the organization, accuracy, and efficiency of clinical records,” a general practitioner writes on Sermo. “Provided that physicians have complete and reliable patient information, AI can serve as a powerful support tool, reducing administrative workload and enabling healthcare professionals to focus more on delivering high-quality patient care.”
AI isn’t a replacement for reading journals, attending conferences and blocking off study time, but the tools can augment your efforts.
Replace your unread bookmark folder with AI summaries
Perhaps you have a folder full of journal articles you fully intended to read and never do. AI tools like OpenEvidence can synthesize recent literature in your specialty into evidence digests you can review in two to three minutes between patients. Instead of one 45-minute reading session you keep postponing, you get several short, AI-assisted touchpoints throughout the day. It’s not a replacement for reading studies, but it can supplement your reading.
Close knowledge gaps in real time
When an unfamiliar condition, drug interaction or treatment protocol comes up mid-encounter, AI clinical decision support tools can offer an evidence-grounded explanation on the spot. This allows you to review unfamiliar topics in the context of a real clinical question, and you can validate the findings rather than starting your research from scratch.
Find related resources and content
When you read a noteworthy journal article, AI can surface more recent research on the same topic or build a list of articles that support or contradict it. That makes it far easier to see where a single study sits within the wider evidence base.
Physicians’ role in shaping AI
Physicians are best positioned to shape how AI tools evolve. The American Medical Association (AMA) has emphasized that physicians should be involved in developing AI training standards, not simply consuming AI tools marketed to them. The hope is that physicians won’t passively consume AI output without critical evaluation and become dependent on tools they don’t fully understand.
Physician involvement can take many forms: serving on institutional AI committees, providing structured feedback on CME platforms or sharing real-world experiences in professional communities. Sermo is one of the few platforms where physicians can discuss AI tools, share experiences with specific products and debate the implications of AI in clinical practice. For example, Sermo members have weighed in on what they think is the biggest missed opportunity for AI in medicine right now: reducing low-value administrative work (24%), giving physicians more time back with patients (21%) and reducing burnout and cognitive load (21%).
AI is already transforming medicine
AI is already embedded in clinical decision support, CME platforms and board prep tools. You may have finished schooling before AI-focused courses existed, but you can use the above strategies to choose a reliable tool and integrate it into your workflow.
These tools can make mistakes, so they aren’t a replacement for your clinical judgment. Ideally, you’ll choose a tool that cites its sources and draws from peer-reviewed literature. As a physician, you can also play an important role in engaging with how AI develops.
If you’d like more advice, join Sermo to discuss AI tools, compare experiences with specific products and get guidance on integrating AI into your practice and continuing education. Members can also earn through paid medical surveys, sharing insights while shaping the future of medicine.
