At the ARVO Annual Meeting, new discussions highlighted how artificial intelligence could transform AMD clinical trials, while raising important questions around standardization, validation and the role of human judgment in patient care.
At a time when ophthalmic clinical trials are becoming increasingly data-rich and imaging-driven, there is growing interest in how AI can improve the precision, scalability and consistency of trial endpoints. AI-based image analysis has shown clear promise in areas such as biomarker quantification, but important challenges remain around standardization, cross-platform reproducibility and defining which AI-derived measures are truly clinically meaningful.
A session at the ARVO Annual Meeting 2026 explored the evolving role of AI in Age-related Macular Degeneration clinical trials, highlighting both the transformative potential of automated imaging analysis and the practical barriers that continue to limit broader implementation. Central to this discussion was the CLEAR study, an initiative led through the International Retinal Imaging Society that aims to establish a standardized validation framework for AI tools used in clinical-trial decision-making.
In this Q&A, Dr Omer Trivizki (Bascom Palmer Eye Institute and Tel Aviv Medical Center) discusses how these same principles of standardization, transparency and clinically meaningful interpretation extend beyond imaging algorithms to emerging large language models. Drawing on his work at the Israeli GA Research Center and through CLEAR, he reflects on the risks of incomplete yet highly convincing AI-generated patient advice, particularly in rare inherited retinal diseases, and why clinical judgment, critical thinking and human connection must remain at the center of patient education and trial decision-making.
Don’t miss out on hearing about our latest peer reviewed articles, expert opinions, conference news, podcasts and more.
Given the significant variability in medical accuracy and tone across models, what is the single most concerning gap you identified in how LLMs counsel patients with rare inherited diseases?
The most concerning gap is not simply that LLMs can be inaccurate. It is that they can be incomplete while sounding very confident, fluent, and reassuring.
In rare inherited diseases, especially inherited retinal diseases, counseling is not just about naming the condition. It is about helping a patient and often an entire family understand what the diagnosis may mean, what testing is needed, whether relatives should be evaluated, what the prognosis is, what trials may be relevant, and what uncertainty still remains. An AI answer may sound kind and polished, but still miss the human and clinical context.
For me, this connects directly to our work in AI validation. At the Israeli GA Research Center, and through the CLEAR study under IntRIS (the International Retinal Imaging Society), we are focused on standardization, transparency, and clinically meaningful interpretation of AI outputs. The same principles should apply to LLMs.
AI can be incredibly helpful, but we must not allow it to replace clinical judgment, critical thinking, or the human ability to understand what a patient is truly asking beneath the question.
How should ophthalmologists adapt their patient education workflows to address the likelihood that patients are receiving inconsistent or incomplete disease management advice from AI chatbots?
Ophthalmologists should start by accepting that many patients are already using AI. The answer is not to dismiss it, but to bring it into the room. I would make it part of the consultation: “Have you looked this up online or asked an AI tool about it?” That simple question can reveal fears, misconceptions, or incomplete advice, and allows us to guide the patient without judgment.
Patient education now needs to be more structured. Patients should leave the clinic with a short, clear summary: what we know, what remains uncertain, what testing is needed, whether family members should be evaluated, when follow-up is important, and which sources are reliable.
This is especially important when discussing clinical trials. AI may help screen eligibility criteria, but that is only one part of the decision. As physicians, we understand what trials are available, but we also assess the whole patient: systemic health, functional status, visual needs, ability to attend visits, family context, expectations, and whether participation is truly appropriate.
AI can help with information and organization, but medicine is not just data matching. Critical thinking, clinical experience, ethics, and human presence will become more important, not less. AI should support the physician, not replace the physician.
Disclosures: Dr Omer Trivizki has nothing to disclose in relation to this article. No fees or funding were associated with this article.
Cite: Omer Trivizki. ARVO 2026: Can AI transform AMD clinical trials while keeping patient care human? touchOPHTHALMOLOGY. 15 May 2026.
Editor: Nicola Cartridge, Head of Content
Acknowledgments: This content has been developed independently by Touch Medical Media for touchOPHTHALMOLOGY. It is not affiliated with ARVO. Views expressed are the speaker’s own and do not necessarily reflect the views of Touch Medical Media.
