“ChatGPT can pass medical exams now – does that mean a bot will soon read every OCT in my clinic?”
If some version of this question has briefly crossed your mind, you’re not alone. Artificial intelligence (AI) is continuously making headlines – claims of algorithms interpreting retinal images better than experts, or large language models (LLMs) drafting accessible clinical letters in seconds. The pace of AI progress feels dizzying. Yet the everyday realities of clinic backlogs, NHS budget cuts, and aging computer systems quickly bring us back to earth.
This blog post unpacks where we stand with practical and safe AI adoption in ophthalmology, why a credible AI directory helps, and where AI innovation may take us next.
Why AI matters to ophthalmologists today
The prospect of AI-enabled ophthalmology services is a present-day reality. While translation of AI algorithms from ‘code to clinic’ has been slow, their potential remains. Waiting lists for ophthalmology appointments continue to grow, as does the amount of admin generated from clinical care. A proliferation of autonomous diabetic retinopathy algorithms promise to screen large volumes of images with 99% sensitivity. Ambient scribes are slowly entering the clinics with the promise of reducing our workload, and patients themselves are asking different questions about their management. As a profession, we need to know more about AI – with knowledge rooted in evidence rather than anecdotes and marketing materials.
The regulatory maze
In order for an AI algorithm to be used in clinical care, it must first receive regulatory approval. Establishing the presence and scope of these approvals can be confusing to navigate for end users like ourselves. Information is split across multiple silos: regulatory details, clinical evidence, and adverse events are spread across different databases, with some key aspects outside of the public domain. To further complicate matters, the same AI product may undergo name changes between jurisdictions and sequential version updates. People often see regulatory approval as a guarantee that a product will be ‘safe and effective’, but this is only one part of the story. Regulators work to ensure that a product performs safely as claimed, but this may not necessarily equate to the real-world performance that adopters are looking for.
Against that backdrop, the College launched a public, vendor‑neutral directory of ophthalmic AI tools that have been approved by regulators and are available for clinical use in the UK. The aim is to support informed exploration, evaluation, and adoption of AI in clinical practice.
What the evidence tells us so far
In parallel, our group has recently published a scoping review of AI tools for ophthalmic image analysis to map the landscape and identify good practice and research gaps. We identified 36 commercially available products from 28 manufacturers in three jurisdictions.(1) Nearly all were cleared in the EU, 22% in Australia and 8% in the United States. 19% had no peer-reviewed performance data in the public domain. Of the 131 evaluation studies retrieved, only 8% were interventional, demographics were poorly reported in validation datasets, and only a third were conducted independently of the manufacturer. The takeaway is clear: the evidence base behind ophthalmic AI tools varies significantly, and buyers should think carefully about how they may perform locally.
Looking ahead: from pilot to practice
AI will not replace ophthalmologists, but it may well contribute to what “good” looks like in eye care. The College’s AI directory and ophthalmologists’ ability to critically appraise existent research are valuable foundations to achieving this, but the real progress will be built by everyday teams who test, audit and – most importantly – share their results. So, the next time you scroll past an eye‑catching AI headline, don’t just share it. Ask: could this solve a real problem in my clinic? What evidence would persuade my governance committee? Who else needs to be in the room?
Send your discoveries, blind alleys and breakthrough moments to [email protected]. Together, we can learn from each other’s successes and failures to ensure that ophthalmology patients across the UK benefit safely from AI innovation.
References:
- Ong AY, Taribagil P, Sevgi M, Kale AU, Dow ER, Macdonald T, et al. A scoping review of artificial intelligence as a medical device for ophthalmic image analysis in Europe, Australia and America. npj Digit Med. 2025 May 29;8(1):1–13.