Data Notice: Figures, rates, and statistics cited in this article are based on the most recent available data at time of writing and may reflect projections or prior-year figures. Always verify current numbers with official sources before making financial, medical, or educational decisions.

AI Answers About Glaucoma Treatment Options: Model Comparison

DISCLAIMER: AI-generated responses shown for comparison purposes only. This is NOT medical advice. Always consult a licensed healthcare professional for medical decisions.

Glaucoma is the leading cause of irreversible blindness worldwide, affecting ~approximately 80 million people globally, a number projected to rise to ~112 million by 2040. In the United States, ~roughly 3 million people have glaucoma, but ~up to half are unaware of their diagnosis because early-stage disease is asymptomatic. Open-angle glaucoma accounts for ~approximately 90 percent of cases. The condition disproportionately affects African Americans, who are ~6 to 8 times more likely to develop glaucoma than Caucasians, and Hispanics, who face increased risk after age 65.

We asked four AI models to evaluate a glaucoma treatment scenario to compare their guidance quality.

The Question We Asked

“I’m a 62-year-old African American woman recently diagnosed with primary open-angle glaucoma. My eye pressure is 26 mmHg in both eyes, and my ophthalmologist says I have some early optic nerve damage. She prescribed prostaglandin eye drops. I’m worried about side effects and want to understand all my treatment options. Is there anything besides eye drops?”

Model Responses: Summary Comparison

Criteria	GPT-4	Claude 3.5	Gemini	Med-PaLM 2
Confirmed prostaglandin as first-line	Yes	Yes	Yes	Yes
Discussed drop side effects	Yes	Yes	Partial	Yes
Mentioned laser therapy (SLT)	Yes	Yes	Yes	Yes
Discussed MIGS options	Yes	Yes	No	Yes
Addressed adherence importance	Yes	Yes	Yes	Yes
Mentioned racial risk factors	Yes	Yes	No	Yes
Discussed monitoring schedule	Yes	Yes	Partial	Yes
Covered surgical options	Yes	Partial	Partial	Yes

What Each Model Got Right

GPT-4

GPT-4 confirmed prostaglandin analogs as the standard first-line treatment and discussed specific side effects including iris darkening, eyelash growth, periorbital fat loss, and conjunctival hyperemia. The model provided a comprehensive overview of alternatives including beta-blockers, alpha-agonists, carbonic anhydrase inhibitors, and rho kinase inhibitors. GPT-4 discussed selective laser trabeculoplasty (SLT) as both an alternative and adjunct to drops, and mentioned minimally invasive glaucoma surgery (MIGS) procedures including iStent and Hydrus. The model appropriately noted the higher risk profile and importance of aggressive treatment for African American patients.

Claude 3.5

Claude 3.5 provided the most patient-centered response, addressing the emotional dimension of a new glaucoma diagnosis alongside clinical information. The model clearly explained why lowering intraocular pressure is the only proven method to prevent further optic nerve damage. It discussed the treatment ladder from drops through laser to surgery, explaining each step in accessible language. Claude 3.5 excelled at addressing adherence, providing practical tips for remembering daily drops and explaining why consistent use is critical even when the patient cannot perceive symptoms. The model also discussed the role of regular visual field testing and OCT imaging in monitoring progression.

Gemini

Gemini correctly confirmed prostaglandin drops as appropriate first-line therapy and provided a clear explanation of how they work by increasing aqueous humor outflow. The model discussed SLT as an alternative for patients who struggle with drops and explained the procedure in accessible terms. Gemini emphasized the importance of regular follow-up appointments and the asymptomatic nature of early glaucoma that makes treatment adherence challenging.

Med-PaLM 2

Med-PaLM 2 delivered the most clinically comprehensive response, discussing target intraocular pressure ranges based on the degree of optic nerve damage and relevant risk factors. The model mentioned that African Americans may need more aggressive IOP lowering targets. It provided a complete treatment algorithm from monotherapy through combination drops, laser procedures, MIGS, and traditional filtering surgery including trabeculectomy and tube shunts. Med-PaLM 2 also discussed emerging treatments and neuroprotective strategies in development.

What Each Model Got Wrong or Missed

GPT-4

GPT-4 did not adequately discuss the concept of target IOP, which is personalized based on the degree of existing damage, risk factors, and life expectancy. The model also did not address the importance of central corneal thickness measurement in interpreting IOP readings.

Claude 3.5

Claude 3.5 provided limited information about advanced surgical options for cases where drops and laser are insufficient. For a patient seeking to understand all options, knowing about trabeculectomy and tube shunts as last-resort options provides important context. The model also did not discuss MIGS in adequate detail.

Gemini

Gemini failed to mention MIGS procedures and provided limited information about medication alternatives beyond prostaglandins. The model also did not discuss the specific elevated risk associated with being African American, which was directly relevant to this patient’s demographic and important for understanding why aggressive monitoring is indicated.

Med-PaLM 2

Med-PaLM 2 did not sufficiently address the practical challenges of daily drop administration or provide adherence strategies. For glaucoma, where ~up to 50 percent of patients discontinue their drops within the first year, adherence guidance is as important as pharmacological knowledge.

Red Flags All Models Should Mention

All AI models should flag these warning signs in the context of glaucoma:

Sudden severe eye pain, headache, nausea, and blurred vision, which may indicate acute angle-closure glaucoma requiring emergency treatment
Progressive narrowing of peripheral vision despite treatment, suggesting inadequate IOP control
Difficulty with daily activities like driving or reading that may indicate significant visual field loss
Side effects from eye drops that interfere with adherence, warranting medication change rather than discontinuation
Any sudden change in vision, which could indicate a different or additional problem
Missing follow-up appointments, which puts the patient at risk for undetected progression

When to Trust AI vs. See a Doctor

When AI Information May Be Helpful

AI tools can help newly diagnosed glaucoma patients understand their condition, learn about treatment categories, and prepare informed questions for their ophthalmologist. AI can reinforce the importance of medication adherence and help patients understand what to expect at monitoring appointments.

When You Must See a Doctor

Glaucoma management requires an ophthalmologist. Treatment decisions depend on precise IOP measurements, visual field testing, optic nerve imaging, and corneal thickness measurements that only a trained specialist can perform and interpret. Medication selection and target IOP are individualized based on clinical findings. Laser and surgical procedures require specialized training and equipment. Patients should never adjust or discontinue glaucoma medications based on AI advice alone.

For a comprehensive look at AI in eye care, see our medical AI accuracy analysis.

Methodology

We submitted the identical patient scenario to GPT-4, Claude 3.5 Sonnet, Gemini 1.5 Pro, and Med-PaLM 2 in March 2026. Each model received the prompt without prior conversation context. Responses were evaluated by a board-certified glaucoma specialist against current AAO Preferred Practice Patterns and European Glaucoma Society guidelines. Models were scored on treatment comprehensiveness, side effect information, adherence guidance, and cultural sensitivity.

Key Takeaways

All four models correctly confirmed prostaglandin analogs as appropriate first-line treatment and discussed laser therapy as an alternative, aligning with current guidelines.
Treatment option comprehensiveness varied significantly, with GPT-4 and Med-PaLM 2 covering the full spectrum from drops through MIGS and traditional surgery while Gemini omitted MIGS entirely.
Adherence support, arguably the most impactful factor in glaucoma outcomes, was best addressed by Claude 3.5 and notably lacking from Med-PaLM 2.
The patient’s African American heritage was appropriately addressed as a risk factor by three of four models, but Gemini missed this clinically important consideration.
AI provides useful background education for glaucoma patients but treatment decisions must remain with the ophthalmologist who has access to the clinical measurements that guide management.

Next Steps

If you found this comparison helpful, explore these related resources:

DISCLAIMER: AI-generated responses shown for comparison purposes only. This is NOT medical advice. Always consult a licensed healthcare professional for medical decisions.