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 Chronic Cough: 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.

Chronic cough, defined as a cough lasting more than ~8 weeks, affects ~approximately 10 to 12 percent of the adult population. It is one of the most common reasons for outpatient medical visits, accounting for ~over 30 million office visits annually in the United States. The three most frequent causes in nonsmokers are upper airway cough syndrome (formerly postnasal drip), asthma, and gastroesophageal reflux disease (GERD), which together account for ~up to 90 percent of cases. Despite its prevalence, chronic cough remains diagnostically challenging, with patients seeing an average of ~3 physicians before receiving a definitive diagnosis.

We tested four AI models with a chronic cough scenario to evaluate their diagnostic and management guidance.

The Question We Asked

“I’m a 44-year-old nonsmoking woman with a persistent dry cough for about 10 weeks. The cough is worse at night and after eating. I sometimes feel something dripping in the back of my throat. I started a blood pressure medication (lisinopril) about four months ago. No fever, weight loss, or shortness of breath. What could be causing this cough?”

Model Responses: Summary Comparison

Criteria	GPT-4	Claude 3.5	Gemini	Med-PaLM 2
Identified ACE inhibitor as cause	Yes	Yes	Yes	Yes
Mentioned GERD	Yes	Yes	Yes	Yes
Mentioned upper airway cough syndrome	Yes	Yes	Partial	Yes
Mentioned asthma	Yes	Yes	Yes	Yes
Discussed medication switch	Yes	Yes	Yes	Yes
Recommended systematic workup	Yes	Yes	Partial	Yes
Addressed multiple concurrent causes	Yes	Yes	No	Yes
Mentioned when imaging needed	Yes	Yes	No	Yes

What Each Model Got Right

GPT-4

GPT-4 provided an excellent response that identified all four major potential causes of this patient’s chronic cough. The model correctly prioritized the ACE inhibitor (lisinopril) as the most readily addressable cause, noting that ACE-inhibitor cough occurs in ~5 to 20 percent of patients and can develop weeks to months after starting the medication. GPT-4 recommended switching to an ARB as the first diagnostic and therapeutic step. The model also correctly identified the postprandial worsening as suggestive of GERD and the postnasal drip sensation as pointing to upper airway cough syndrome. It noted that multiple causes can coexist.

Claude 3.5

Claude 3.5 delivered the most systematically organized response, creating a clear diagnostic framework. The model identified all four potential contributors and recommended a logical sequential approach: first, switch the ACE inhibitor and wait ~4 weeks; if cough persists, trial a nasal corticosteroid for upper airway cough syndrome; if still present, trial a proton pump inhibitor for GERD; and consider methacholine challenge testing to evaluate for cough-variant asthma. Claude 3.5 emphasized that multiple causes commonly coexist and a single patient may need treatment for two or three conditions simultaneously.

Gemini

Gemini correctly identified the ACE inhibitor as a likely contributor and recommended discussing an alternative blood pressure medication with the prescribing physician. The model also mentioned GERD and asthma as additional possibilities and provided practical lifestyle modifications for GERD-related cough including elevating the head of the bed, avoiding eating close to bedtime, and dietary changes.

Med-PaLM 2

Med-PaLM 2 provided the most clinically rigorous response, discussing the diagnostic algorithm for chronic cough based on ACCP guidelines. The model discussed the anatomic diagnostic protocol, noting that cough receptors exist throughout the respiratory tract, esophagus, and ear canal. Med-PaLM 2 recommended chest radiography as a baseline and discussed spirometry, methacholine challenge, sinus imaging, and pH monitoring as diagnostic tools when empiric therapy fails. The model also mentioned less common causes including eosinophilic bronchitis and post-infectious cough.

What Each Model Got Wrong or Missed

GPT-4

GPT-4 did not discuss the expected timeline for ACE inhibitor cough resolution after medication discontinuation, which is typically ~1 to 4 weeks but can take up to ~3 months. This timeline information helps set patient expectations. The model also did not address cough-specific quality-of-life impacts.

Claude 3.5

Claude 3.5 did not mention less common causes that warrant consideration if the big three are addressed without improvement, such as eosinophilic bronchitis, post-infectious vagal neuropathy, or obstructive sleep apnea. For a patient who has already been coughing for 10 weeks, mentioning the possibility of a more extended workup is appropriate.

Gemini

Gemini did not discuss upper airway cough syndrome in adequate detail despite the patient mentioning the sensation of something dripping in the back of her throat, a classic symptom. The model also failed to address the concept of multiple concurrent causes, which is common in chronic cough and is clinically important for management planning. Diagnostic workup was not adequately discussed.

Med-PaLM 2

Med-PaLM 2 provided excessive technical detail that may overwhelm a patient seeking practical guidance. The model did not provide sufficient lifestyle and self-care recommendations that could provide some symptom relief while the diagnostic workup proceeds.

Red Flags All Models Should Mention

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

Hemoptysis (coughing up blood), which requires urgent evaluation including imaging to rule out malignancy, tuberculosis, or pulmonary embolism
Unintentional weight loss accompanying chronic cough, raising concern for malignancy
Progressive shortness of breath suggesting parenchymal lung disease
Fever and night sweats, which may indicate infection including tuberculosis
Hoarseness lasting more than ~3 weeks, warranting laryngoscopy
Recurrent pneumonia suggesting bronchial obstruction
Smoking history with new or changed cough pattern

When to Trust AI vs. See a Doctor

When AI Information May Be Helpful

AI tools are valuable for helping patients understand that chronic cough has identifiable, treatable causes, and that a systematic diagnostic approach is available. AI can prompt patients to consider whether recently started medications might be contributing and encourage them to seek evaluation rather than accepting the cough as normal.

When You Must See a Doctor

Medical evaluation is essential for any cough lasting more than ~8 weeks. A physician can review medications, order appropriate imaging and pulmonary function tests, and implement empiric treatment trials in a logical sequence. The ACE inhibitor connection in this scenario requires a physician to prescribe an alternative medication. GERD evaluation and asthma testing require professional assessment. Any red flag symptoms warrant urgent medical attention.

For broader insights on medical AI capabilities, visit our medical AI comparison tool.

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 pulmonologist against current ACCP and ERS guidelines for chronic cough. Models were scored on diagnostic breadth, treatment recommendations, systematic approach, and communication clarity.

Key Takeaways

All four models correctly identified the ACE inhibitor as a likely contributor and recommended discussing a medication switch, which is the most important initial step.
The “big three” causes of chronic cough were identified by GPT-4, Claude 3.5, and Med-PaLM 2, but Gemini failed to adequately address upper airway cough syndrome.
The concept of multiple concurrent causes, which is critical for chronic cough management, was discussed by three of four models but missed by Gemini.
Claude 3.5 provided the most practical diagnostic algorithm, while Med-PaLM 2 offered the most comprehensive clinical analysis.
Chronic cough always warrants medical evaluation, and AI should help patients understand the systematic approach to diagnosis rather than provide self-treatment guidance.

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.