will AI replace optometrists?

Every single core task in optometry sits at 0% AI penetration. That's not a fluke. It reflects something real: this job is built around a licensed clinician physically examining a patient, making judgment calls, and taking legal responsibility for what happens next. When you use a slit lamp to check for early signs of glaucoma, or dilate someone's pupils to inspect the retina, you're doing something that requires your hands, your training, and your presence in the room.

The diagnostic side is where your judgment is hardest to replace. You're not just reading numbers off a chart. You're integrating visual acuity data, patient history, symptoms, medication interactions, and the appearance of the optic nerve to build a picture. If a patient's field test looks suspicious, you decide whether that's a calibration issue, patient fatigue, or early glaucomatous change. That call has consequences, and the liability sits with you.

The patient relationship matters too. Educating someone on why they keep getting eye infections from their contact lens habits, or talking a nervous patient through what cataract surgery recovery looks like, requires reading a person in real time. You adjust your language, your tone, your level of detail based on who's sitting across from you. And the procedural tasks, removing a foreign body from someone's eye, fitting a scleral lens for a patient with keratoconus, managing post-surgical care, are completely physical. There's no remote or automated version of that.

AI image analysis tools like IDx-DR can screen retinal photos for diabetic retinopathy, and they do that one narrow task reasonably well. But screening is not diagnosing. When the AI flags something, a licensed optometrist still has to examine the patient, interpret the result in context, and decide what to do. The AI has no access to the rest of the clinical picture.

Hallucination is a serious problem in any AI system used near clinical data. Large language models asked to summarise patient records or suggest treatment options will sometimes produce plausible-sounding but wrong information. In optometry, that could mean a missed drug interaction in a patient already on systemic medications, or a misread of a visual field result. There's no regulatory framework that currently allows an AI to write a prescription for corrective lenses or ophthalmic medications, let alone be held accountable if something goes wrong.

Privacy is the other gap. Patient data in optometry is protected under HIPAA, and feeding clinical notes or imaging into a general-purpose AI tool creates real legal exposure. Most practice management systems haven't solved this cleanly. Until there's a clear compliance pathway, many of the AI tools that look promising on paper can't be used in a clinical setting without significant legal risk.

The AI tools making the most real-world difference in optometry right now are in screening and imaging analysis, not clinical decision-making. IDx-DR, now marketed as Digital Screening for DR by Digital Diagnostics, is an FDA-cleared tool that analyzes retinal photographs for signs of diabetic retinopathy without a clinician reading the image first. It's used in primary care settings to flag patients who need a referral. You might encounter patients who've already been screened this way before they walk into your chair.

Notal Vision's remote monitoring platform lets patients with age-related macular degeneration test their own vision at home using a device that uploads data to a cloud-based AI for monitoring between clinic visits. This expands your ability to track a condition over time without requiring the patient to come in for every check. It doesn't replace your assessment; it gives you more data points.

On the practice management side, tools like Eyefinity and RevolutionEHR have started integrating AI-assisted coding suggestions, which can speed up billing and reduce claim rejections. Some practices are also using AI scribing tools like Suki to capture exam notes by voice during the appointment. These tools handle the administrative side of the encounter, the note structure, the billing codes, the referral letter drafts, not the clinical side. That distinction matters. The documentation help is real and saves time. The clinical AI is still narrow and supervised.

The biggest shift is in where your time goes after the exam. If your practice uses an AI scribing tool, you're spending less time typing up the encounter note and more time on the actual patient conversation. The exam itself hasn't changed at all. You're still behind the slit lamp, still doing the same tests, still making the same calls.

Remote monitoring platforms mean some of your follow-up work now happens asynchronously. A macular degeneration patient might send in home test data three times a week, and you review a summary rather than waiting for a six-week appointment. Your schedule can shift to handle more new patients or complex cases because routine monitoring is partially offloaded.

What hasn't changed at all is the exam room itself. The refraction, the dilation, the contact lens fitting, the pressure check, the conversation about symptoms: that's all still you, in person, in real time. Admin got a little faster in practices that have adopted these tools. The clinical work is exactly what it was.

The BLS projects 8% growth for optometrists between 2024 and 2034, faster than the average for all occupations. With 47,800 optometrists employed in 2024 and around 2,400 openings per year, the math is straightforward: demand is outpacing supply, and AI isn't filling the gap.

The growth is driven by demographics. The U.S. population over 65 is growing, and older adults need more eye care. Conditions like glaucoma, macular degeneration, and diabetic retinopathy are all age-linked, and all require ongoing licensed clinical management. AI screening tools may catch more cases earlier, but they generate referrals, which means more patients who need to see you, not fewer.

There's also a geographic shortage that AI can't solve. Rural and underserved areas have too few optometrists relative to need. Tele-optometry is expanding slightly, but you still need a licensed provider to interpret results and prescribe. The Anthropic Economic Index rates optometry at essentially 0% AI exposure across core tasks, and the BLS growth projection reflects a profession where demand is structural, not discretionary. These aren't jobs that disappear when a company cuts its software budget.

The 0% AI exposure score for optometry is likely to hold for the next five to ten years. The tasks that define the role are physical, licensed, and legally accountable in ways that are genuinely hard for AI to replace. Diagnostic imaging AI will keep improving, and tools like those for diabetic retinopathy screening will probably become standard in more clinical settings. But better screening creates more referrals, which means more demand for your time, not less.

For AI to seriously threaten the optometrist role, you'd need autonomous refraction that's legally recognized as a prescription, AI that can physically examine an eye and remove a foreign body, and regulatory approval for AI to carry the liability currently held by a licensed clinician. None of that is close. The FDA has been cautious about autonomous AI diagnostics even in narrower domains. A 10-year horizon doesn't change that picture much. The procedural and relational core of this job is structurally resistant to automation in a way that, say, a data processing role isn't.

The clearest thing to double down on is the diagnostic depth that makes you irreplaceable. Complex cases, systemic disease management (diabetic eye disease, hypertensive retinopathy), and co-management with ophthalmology are where your judgment has the most value. If you're building a career now, getting comfortable with interpreting AI-assisted imaging results is worth doing, not because AI will take over, but because patients will arrive having already been screened, and you'll need to contextualise those results.

The administrative tools covered above are worth adopting if you haven't. Practices that use AI-assisted documentation and billing are genuinely more efficient, and efficiency means you can see more patients or finish the day faster. Neither outcome hurts your career.

The skills that matter most long-term are the ones with no AI analogue: procedural competence (foreign body removal, contact lens fitting for irregular corneas, pre- and post-op care), the ability to spot systemic disease through the eye, and the clinical relationship that keeps patients coming back. Specialising in low vision, pediatric optometry, or sports vision gives you a moat that a general screening algorithm can't touch. The profession is growing, the exposure to AI displacement is as low as it gets, and the best career move is to go deeper into clinical complexity rather than worrying about being replaced by software.