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 Conjunctivitis Types: 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.

Conjunctivitis, commonly known as pink eye, is one of the most frequent eye conditions, accounting for ~approximately 1 percent of all primary care visits and ~up to 30 percent of all eye-related complaints. Viral conjunctivitis is the most common type in adults, causing ~up to 80 percent of infectious conjunctivitis cases. Bacterial conjunctivitis is more common in children and accounts for ~50 to 75 percent of pediatric cases. Allergic conjunctivitis affects ~up to 40 percent of the population at some point, particularly in regions with high pollen counts. Each type requires different management, making accurate differentiation clinically important.

We asked four AI models to differentiate conjunctivitis types based on a patient scenario.

The Question We Asked

“I’m a 33-year-old man and I woke up with my left eye crusted shut and very red. There’s a thick yellowish-green discharge, and the eye feels gritty but not particularly itchy. My right eye is starting to get slightly red too. I don’t have allergies and I haven’t been around anyone with pink eye that I know of. How do I know if this is bacterial, viral, or allergic, and does it matter for treatment?”

Model Responses: Summary Comparison

Criteria	GPT-4	Claude 3.5	Gemini	Med-PaLM 2
Correct likely diagnosis	Yes	Yes	Yes	Yes
Distinguished all three types	Yes	Yes	Partial	Yes
Identified discharge type significance	Yes	Yes	Yes	Yes
Discussed antibiotic treatment	Yes	Yes	Yes	Yes
Addressed contagion precautions	Yes	Yes	Yes	Yes
Mentioned when to see a doctor	Yes	Yes	Yes	Yes
Discussed differential diagnoses	Yes	Partial	No	Yes
Addressed contact lens considerations	Yes	Yes	No	Partial

What Each Model Got Right

GPT-4

GPT-4 correctly identified the presentation as most consistent with bacterial conjunctivitis based on the purulent discharge, morning crusting, and gritty sensation. The model provided an excellent comparison table distinguishing viral, bacterial, and allergic conjunctivitis by discharge type, associated symptoms, and typical course. It recommended medical evaluation for antibiotic eye drops, specifically mentioning fluoroquinolone drops as a common first-line choice. GPT-4 discussed hygiene measures including hand washing, avoiding touching the eyes, and not sharing towels or pillows.

Claude 3.5

Claude 3.5 provided the clearest patient-facing explanation of the three conjunctivitis types. The model organized its response around the key differentiating features: discharge quality (purulent in bacterial, watery in viral, stringy and clear in allergic), itching prominence (minimal in bacterial, moderate in viral, severe in allergic), and typical course. Claude 3.5 correctly assessed the described symptoms as most consistent with bacterial conjunctivitis and recommended seeing a healthcare provider for antibiotic drops. The model also provided practical hygiene instructions and addressed when to return to work.

Gemini

Gemini correctly identified bacterial conjunctivitis based on the discharge description and provided clear treatment guidance. The model was effective at explaining in simple terms why the type of conjunctivitis matters for treatment, noting that antibiotics work for bacterial but not viral or allergic forms. Gemini provided strong hygiene and contagion prevention advice, including recommendation to discard eye makeup and replace contact lens cases.

Med-PaLM 2

Med-PaLM 2 provided the most clinically sophisticated analysis, discussing the sensitivity and specificity of clinical signs in distinguishing conjunctivitis types. The model noted that while purulent discharge strongly suggests bacterial etiology, culture may be warranted in cases unresponsive to empiric therapy. Med-PaLM 2 also discussed hyperacute bacterial conjunctivitis caused by Neisseria gonorrhoeae as a sight-threatening entity requiring emergency treatment, and mentioned chlamydial conjunctivitis as a consideration in sexually active adults.

What Each Model Got Wrong or Missed

GPT-4

GPT-4 did not discuss the possibility of chlamydial conjunctivitis, which is relevant in a sexually active adult with unilateral-onset conjunctivitis. The model also did not address the natural history of untreated bacterial conjunctivitis, which often self-resolves in ~5 to 7 days, information relevant to shared decision-making about antibiotic use.

Claude 3.5

Claude 3.5 provided limited differential diagnosis beyond the three conjunctivitis types. Conditions like anterior uveitis, keratitis, and acute angle-closure glaucoma can present with a red eye and require different management. The model should have mentioned warning signs that distinguish these more serious conditions.

Gemini

Gemini did not adequately distinguish all three conjunctivitis types in its response, focusing primarily on bacterial versus viral and giving allergic conjunctivitis insufficient attention. The model also did not mention more serious red eye conditions that should be considered in the differential diagnosis.

Med-PaLM 2

Med-PaLM 2 did not provide sufficient practical self-care guidance. While clinically comprehensive, the model assumed a level of medical sophistication that most patients would not have. Practical information about when to return to work, how to manage symptoms at home while awaiting the doctor visit, and what over-the-counter relief options are available would have improved the response.

Red Flags All Models Should Mention

All AI models should flag these warning signs that suggest a more serious eye condition than simple conjunctivitis:

Significant eye pain (beyond mild grittiness) suggesting keratitis, uveitis, or glaucoma
Decreased vision or visual blurring, which is unusual for simple conjunctivitis
Photophobia (light sensitivity) suggesting anterior uveitis or keratitis
Copious hyperacute purulent discharge developing within ~12 to 24 hours, suggesting gonococcal infection requiring emergency treatment
A fixed, mid-dilated pupil with severe pain, raising concern for acute angle-closure glaucoma
Contact lens wearer with red eye, increasing risk of corneal ulcer
History of recent eye surgery or eye trauma

When to Trust AI vs. See a Doctor

When AI Information May Be Helpful

AI tools can help patients understand the differences between conjunctivitis types, recognize that purulent discharge typically indicates bacterial infection, and implement appropriate hygiene measures to prevent spreading the infection. AI can also help patients determine whether their symptoms warrant urgent versus routine evaluation.

When You Must See a Doctor

A healthcare provider should evaluate any red eye that involves significant pain, visual changes, photophobia, or contact lens wear. While mild viral conjunctivitis may be self-limiting, the patient in this scenario has symptoms consistent with bacterial conjunctivitis that would benefit from antibiotic treatment to speed resolution and reduce contagion. Only a clinician can properly distinguish between conjunctivitis types and rule out more serious conditions through examination.

For more on navigating medical AI tools, see how to ask AI health questions safely.

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 ophthalmologist against current AAO guidelines. Models were scored on diagnostic differentiation, treatment appropriateness, safety warning completeness, and communication clarity.

Key Takeaways

All four models correctly identified the presentation as most consistent with bacterial conjunctivitis based on the purulent discharge pattern.
The ability to distinguish between viral, bacterial, and allergic conjunctivitis was strongest in GPT-4 and Claude 3.5, which provided clear comparative frameworks.
Med-PaLM 2 uniquely raised important considerations about gonococcal and chlamydial conjunctivitis, which the other models missed.
Warning signs for more serious eye conditions were inadequately covered by Claude 3.5 and Gemini, who focused narrowly on conjunctivitis without discussing the broader differential.
AI tools can help with initial symptom assessment but professional evaluation is recommended for any red eye with purulent discharge, pain, or vision changes.

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.