AI and eye health in summary:
- Patients increasingly arrive at my clinic having consulted AI about their eye health.
- Used well, AI is the best patient-preparation tool we have ever had. Used badly, it can mislead – particularly when used for image-based self-diagnosis, where the limitations of a smartphone photograph make reliable assessment impossible.
- Use AI to understand concepts and prepare questions, not to diagnose yourself. Bring what you have read to a specialist, and be open to the answer being different.
AI and your eye health: an ophthalmologist’s view from the consulting room
It is now unusual for me to see a new patient who has not first consulted an AI tool. Patients arrive with summaries of their symptoms, suggested diagnoses, comparisons of treatments and, frequently, specific questions about which procedure they should undergo. As an ophthalmologist, my view of this trend is more nuanced than many might expect.
AI in healthcare is here. The question is no longer whether patients will use it, but how well they will use it – and how clinicians like me will adapt to the more informed, more anxious and, occasionally, the more confused patients that these tools produce. After observing this shift in the consulting room, here is what I think patients and the profession alike need to understand.
What AI is doing well when it comes to eye health
Used carefully, AI is the best patient preparation tool we have ever had. Patients arrive understanding the basics of what their condition might be, what their treatment options are, and what questions to ask.
For descriptions of common eye conditions, high-level explanations of treatment categories, and guidance on when to seek specialist help, current AI tools are, on the whole, generally reliable. In particular, patients with rare or unfamiliar conditions can now access information they could not have easily found ten years ago. The democratisation of medical literacy is real, and it is welcome in my clinic.
Where the picture becomes more complicated when it comes to AI and eye health
The concerns are not with patients using AI – they are with how AI is being used. There are four patterns I am seeing in my clinic that warrant attention.
Image-based self-diagnosis
Patients are increasingly uploading photographs of their eyes to AI tools and asking for a diagnosis. This is the area I am most concerned about. The eye is a three-dimensional structure with multiple layers. A photograph of the front of the eye, taken on a smartphone in poor lighting, simply cannot capture this. The equivalent would be trying to assess a building’s structural integrity from a photograph of its front door.
The problem is not academic. A red eye, for example, could be conjunctivitis (benign and self-limiting), uveitis (urgent specialist treatment required), a corneal ulcer (potentially sight-threatening), or acute angle-closure glaucoma (a medical emergency).
AI hallucination – when confident is not the same as correct
AI is designed to sound authoritative. It is not designed to admit it does not know and is often overly agreeable. The result is what the field calls “hallucination” – confidently delivered, plausible-sounding medical content that is simply wrong. The confident voice of an AI tool is not the same as a medically-correct voice.
The personalisation problem
Ophthalmology is intensely personal. The right treatment depends on a patient’s prescription, various measurements of their eyes, their age, lifestyle, previous eye health, systemic health and surgical aims. AI cannot perform corneal topography (the detailed mapping of corneal shape), biometry (the precise measurement of the eye’s dimensions), or a slit lamp examination (the magnified examination of the eye’s internal structures). It doesn’t know exactly what a particular surgery can and cannot achieve relative to a patient’s individual situation.
Anxiety amplification
AI tools tend to present worst-case scenarios alongside benign explanations. A patient who experiences floaters in their vision – common, usually harmless – will be told about retinal detachment, another but much more serious cause of floaters in the same response. The lived experience of the patient who arrives in clinic, convinced their retina is detaching is materially different. The consequences of that anxiety – rushed decisions, lost sleep, avoidance of activities – are real, and could have been avoided in the vast majority of cases if they had been able to speak to an eye care professional before their appointment rather than try to self-diagnose.
How to use AI well as a patient
My advice to patients using AI tools is straightforward.
First, use AI to understand concepts, not to make diagnoses. Reading about what laser eye surgery involves before your consultation is genuinely useful. Asking AI which surgical technique you should have, is not.
Second, do not upload photographs of your eye to AI tools and act on the response. There is no current consumer AI tool that can reliably diagnose an eye condition from a photograph.
Third, bring your AI-generated questions to your specialist – but be open to the answer being different. Often the most important thing I do in a consultation is recontextualise: the procedure the patient has read about is real, but it is not the right procedure for them. That conversation needs to happen in a clinic, not on a screen.
Fourth, treat confident AI claims about specific drugs, procedures or statistics with caution. Verify these with a specialist before acting on them.
The future: AI as part of specialist care
The future of eye health is human and AI working together, not one or the other. Patients who arrive well-informed are welcomed in my clinic. The patient I most need to spend additional time with, is not the one who has used AI, but the one who has used it and become convinced they already know what they need.
Use AI as a starting point. Use it to understand, to prepare, and to formulate questions. Then bring those questions to a specialist who can examine your eye, measure it accurately, and help you make sense of what AI alone cannot see.
Furthermore, AI is already integrated into my clinic. For example, some of the scanners have AI assisted data interpretation. AI will transcribe your consultation with me, to help draft the clinic letter that will be sent to you after your consultation that summarises all that we discussed.
About Mr Alex Day
Mr Alex Day FRCOphth PhD CertLRS FWCRS is a Consultant Ophthalmic Surgeon at Moorfields Private, with decades of clinical experience in cataract surgery, laser vision correction, refractive lens exchange and Implantable Collamer Lens (ICL) surgery. The author of over 70 peer-reviewed publications, he is widely regarded as the surgeon of choice for medical practitioners and their families seeking vision correction in London.
