AI is no longer new to eye care. Most of us have seen the demos, sat through the panels — including our own at Vision Expo West — and, hopefully, piloted at least one tool. As we enter 2026, the question is no longer whether AI belongs in eye care. It’s which AI myths no longer do.
Myth #1: AI Needs to Be Autonomous to Be Useful
There’s a quiet belief that AI only becomes meaningful when it acts on its own—diagnosing, deciding and operating independently of clinicians. In reality, that framing has slowed adoption more than it has accelerated it.
The AI systems that are actually getting used do something far less dramatic. They help with triage; they surface patterns; they summarize, flag and prioritize. In other words, they assist.
That distinction matters. In clinical care, autonomy is not the default state—judgment is. The most valuable AI tools respect that reality. They make clinicians faster and more consistent without asking them to surrender responsibility.
This is why I’ve always preferred the term augmented intelligence. Not because it’s softer, but because it’s more accurate. AI works best when it reduces cognitive load, not when it tries to replace it.
Autonomy makes for good headlines. Augmentation makes for better care.
Myth #2: More Data Automatically Leads to Better Models
Everyone agrees data matters. Where we get sloppy is assuming that more of it is always better.
In eye care, the problem is rarely a lack of data. It’s the nature of the data we already have—how it’s captured, labeled and contextualized. Models trained on noisy, biased or inconsistent inputs don’t improve just because the dataset gets larger. They often become more confidently wrong.
Some of the most meaningful gains in AI performance haven’t come from scale, but from discipline: tighter definitions, better annotations and closer collaboration between clinicians and technical teams. Quality beats volume—especially in medicine.
In 2026, the advantage won’t belong to whoever has the biggest dataset. It will belong to those who understand their data—and its limitations—the best.
Myth #3: Regulation Will Be the Main Force That Determines Adoption
For years, regulation has been framed as the central bottleneck for AI in health care. That concern hasn’t disappeared, but I think it’s starting to miss the larger dynamic.
A more useful analogy may be the way ridesharing entered our lives. Uber didn’t win because regulators first created clean, predictable rules. It won because people got comfortable using it. Regulation followed behavior.
We may be seeing something similar with generative AI. Clinicians and patients are already using these tools in their personal lives—to write, summarize, explain and organize. As that comfort grows, the perceived risk of AI assistance in professional settings quietly declines.
This doesn’t mean regulation goes away, or that anything should be allowed. But it does suggest that adoption will be shaped less by formal permission and more by practical acceptance. Tools that assist, integrate cleanly and stay in the background will move forward. Tools that add friction, even if technically compliant, will struggle.
The harder problem to solve now isn’t regulatory approval. It’s workflow.
Releasing the Myths
AI in eye care doesn’t need to be autonomous, massive or perfectly regulated to be useful. It needs to fit. It needs to earn trust. And it needs to make the work a little easier for the people doing it.
If we can let go of these myths, 2026 may finally be the year AI becomes less of a talking point—and more of a tool.
If we can let go of these myths, AI stops being a topic of conversation in 2026 and starts being judged the way any clinical tool should be: by whether it works.
Author
