Prompt Engineering for Organic Chemistry Education: A Digital Loom-Based Approach

The integration of generative artificial intelligence systems in higher education has created debates about their teaching effectiveness, especially in subjects that need specific scientific formalism, like organic chemistry. This work investigates how different complexity levels in prompt structure influence the quality of generative AI system responses in specific organic chemistry topics. Using the digital loom analogy as a theoretical framework, we compared the performance of two AI systems (ChatGPT 5 and GPT IQQO assistant) in tasks involving the Cannizzaro reaction and structure-reactivity relationship of the guanidine group. Through content analysis of 494 coded segments, we found that both systems showed similar performance (92.8% vs. 93.1% conceptual adequacy), regardless of prompt complexity. The "orbital interactions" category emerged as the most vulnerable area (88.0% adequacy), revealing limitations shared between AIs and students. Computational vision analysis of 117 meaning units showed complementarity between systems, while scientific image generation revealed critical tension between visual appeal and conceptual rigor. Results indicate that user competence in prompt construction is more important than AI architecture, establishing practical guidelines for effective integration of these technologies in chemistry education.