Why the exam stays in the chair
Will optometrists be replaced by AI? Not on the evidence available today. An eye exam is a run of physical steps with a patient sitting in front of you: dilating a pupil, aiming a slit lamp, swapping lenses, watching how the eye responds, asking which view looks sharper. Software reads an image well. It does not run that room.
Two parts of the job show why. Refraction and prescribing corrective lenses depend on what the person in the chair reports from one lens to the next, and on catching the moment the answers stop making sense. Examining the eye for glaucoma, diabetic damage or a retinal detachment ends in a judgment call: treat, monitor, or send to a surgeon the same day. A license and a liability sit behind that call.
The market backdrop is steady rather than shaky. About 42,790 optometrists work in the United States, median pay is $136,570, and employment is projected to grow 9.5% from 2025 to 2035 (BLS, 2025). Pressure in this job shows up as tasks shifting, not posts vanishing.
What software does, what it assists, what a person keeps
AI is strongest when the input is already an image or a number. Grading retinal photographs for signs of diabetic eye disease, flagging an odd optic nerve on an OCT scan, and sorting screening results by urgency can run start to finish once the capture is done. Share of task time in that group: 0%.
Assistance is a different mode. Here the software drafts and the clinician decides: pulling up prior measurements for comparison, writing the first version of a chart note, highlighting a change between this year’s scan and last year’s, prepping a referral letter. Share of task time where AI supports rather than finishes the work: 22%.
The rest stays with the optometrist. That means hands-on measurement, fitting and checking contact lenses on a live eye, explaining a diagnosis to someone who just learned their vision is changing, and coordinating care with a surgeon or a primary doctor. Share of task time in that group: 78%. Our answer to “Can AI do it?” reads 16 out of 100, and the coverage method page explains how that share of task time is built.
What has actually been tested
Our evidence grade for “Is it better than a person?” is D. A D grade means there is no direct, measured comparison of AI against licensed optometrists doing this job, so we publish no parity number for it. Research on image reading is not the same thing as a test of the whole role.
What would settle it is specific. First, a prospective study that runs an autonomous system and qualified optometrists through the same complete exams, including refraction, prescription, disease detection and the referral decision, and then tracks what happened to the patients. Second, published audits of automated screening used in everyday primary eyecare, reporting the cases it missed and the ones it sent on unnecessarily. Screening tools built for one disease answer a narrow question; a general exam asks many at once. The quality parity method sets out what counts as a measured comparison, and the wider scoring method covers the rest.
When the picture could change
Most likely after 2042 (8 in 10 of our scenarios). The replacement year method explains what that window is measuring.
Two things could pull it earlier. The cost gap on this page is wide, and cheap screening makes it tempting to push more of the capture work to technicians or kiosks with a clinician reviewing remotely. Tele-optometry extends that: images and measurements taken in one place, read in another, which thins out some chair time per patient.
Two things hold it back. A large part of this job is physical, and the robotics needed to do it unsupervised sit at the dexterous humanoid tier shown above, which is not a product you buy today. Licensing and prescribing rules are the other brake: in US practice, a prescription and a treatment decision belong to a licensed clinician who answers for the outcome.
Good to know: automated screening spreading into pharmacies and primary care often sends more people to an optometrist, not fewer.
How optometrists stay needed
Lean into the work that needs a person in the room. Contact lens fitting and follow-up, where comfort and corneal health depend on seeing the lens on the eye. The difficult conversation: explaining progressive vision loss, driving limits, or why surgery is now on the table. And care coordination across an ophthalmologist, a diabetes team and a school or employer.
Two skills raise your value either way. Learn to audit an AI reading rather than accept it: know the failure modes, check the image quality, and record why you agreed or overruled. Then build pediatric, low-vision or disease-management depth, because narrow screening tools do not cover complex cases.
If you are weighing adjacent paths, compare the work rather than the label. Ophthalmologists add surgery and a longer training route. Orthoptists focus on eye movement and binocular vision. Dispensing opticians sit closer to fitting and retail, with a different exposure profile. You can put any two side by side on the job comparison tool.
For wider context, this role sits in the diagnosing and treating practitioners family and the healthcare sector, where hands-on and licensed work behaves differently from desk work. To see how eyecare roles line up against everything else, browse the jobs that most need a person or search the full job rankings.