ESCRS 2026: AI is already improving IOL formulas, but caution is still needed in clinical decision-making

In an interview at the 44th Congress of the European Society of Cataract and Refractive Surgeons (ESCRS), held September 11–15, 2026, in London, Adam Muzychuk, MD, BSc, FRCSC, of the University of Calgary, discussed where artificial intelligence (AI) is already improving ophthalmic care, and where he believes it still needs human oversight.

Where AI is already helping

Muzychuk pointed to the Kane formula, an IOL power calculation formula built on machine learning, as the moment he first noticed AI making a measurable difference. “That was the first time I really felt outcomes improved on the basis of this technology,” he said. He also expects new AI-driven software suites to help with surgical planning going forward.

The limits of relying on AI

Despite that promise, Muzychuk cautioned against leaning on AI too heavily for clinical decision-making. “What gives you the best numbers mathematically may not always be what's best for the patient,” he said, adding that medicine will always retain “an art to it” that a human needs to handle, at least for now. He noted that clinicians' knowledge is typically anchored to what they learned in residency, and that AI could help keep that knowledge current by surfacing the most up-to-date data and recommendations. “AI will be cold and calculating,” he said. “It will know what the best outcomes are for patients, mathematically.”

Misinformation, manipulation, and bias

Muzychuk also addressed patients' growing use of AI to research their own care, calling it a positive development overall, but one with real risks attached. Large language models can hallucinate, he said, and it is not always clear how a model arrived at a given recommendation, particularly for location-specific information. He also warned that AI-generated recommendations, such as rankings of top providers, are vulnerable to manipulation through search engine optimisation. “It's going to come down to who can play the game and train the AI the fastest,” he said, adding that this dynamic will introduce bias that clinicians and patients alike should stay alert to.