Generative artificial intelligence can now produce fluent drug-information responses, therapeutic recommendations, explanatory prose, and answers to professional examination questions. This capability creates a specific problem for pharmacy education: a correct or persuasive response no longer establishes that the learner independently retrieved appropriate evidence, verified its authenticity and relevance, understood the pharmacological mechanism, integrated patient-specific information, recognized uncertainty, or assumed responsibility for the recommendation. The educational question is therefore not whether generative artificial intelligence should be prohibited, but which components of clinical reasoning must remain demonstrably attributable to the learner when machine assistance is available. Current evidence supports a differentiated response. Large language models can perform strongly on selected knowledge benchmarks and can assist with medication information and educational tasks, yet their performance varies across task type, clinical complexity, prompt formulation, source generation, and patient-specific decision requirements. These limitations become particularly important when pharmacy decisions depend on causal pharmacology rather than factual retrieval alone. Verification should consequently be treated as observable intellectual work rather than an instruction appended to an artificial-intelligence-generated answer. Pharmacy assessment should preserve opportunities to determine whether learners can reconstruct the evidence-to-decision pathway, explain pharmacological causation, respond to clinically meaningful case perturbations, identify unsupported sources, and state what remains uncertain. Generative artificial intelligence may appropriately augment pharmacy education, but professional accountability requires that consequential reasoning remain inspectable and attributable.
