My group used ChatGPT, Gemini, and Microsoft Copilot to test three stages of prompts: original, refined, and contextual for the issue of climate-responsive architecture in Malaysia. The prompts became more specific from general hot-climate strategies to practical university-building solutions, then finally to Malaysia-specific campus strategies.
For the original prompt, Microsoft Copilot performed slightly better because it gave not only strategies, but also benefits, examples, and trade-offs. ChatGPT was the clearest and easiest to use for presentation wording, while Gemini was useful for general explanation. However, all three outputs were still too broad because the prompt did not mention Malaysia, university buildings, or specific conditions.
Best AI for Prompt 1: Microsoft Copilot
For the refined prompt, Gemini performed better because it gave the most detailed response to the hot and humid climate. It explained practical strategies such as shading, hybrid ventilation, roof treatment, material selection, and energy efficiency in a more complete way. ChatGPT was useful for design wording and concept explanation, while Copilot was useful for organizing the response into clear categories.
Best AI for Prompt 2: Gemini
For the contextual prompt, Gemini performed the best overall because it gave the strongest location-specific response. It understood Malaysian hot and humid climate more clearly and connected the strategies to cooling load reduction and student comfort. ChatGPT was also strong because it produced more design-oriented ideas for campus spaces, but Gemini was more complete in technical and contextual accuracy. Copilot was useful but less developed.
Best AI for Prompt 3: Gemini
Overall, Gemini performed the best for this assignment because it gave the strongest balance of contextual accuracy, technical depth, and climate-specific reasoning. It was especially useful in the refined and contextual prompts, where the focus shifted to hot and humid climates and Malaysian university campuses.
ChatGPT was the second strongest tool because it produced clearer design explanations and more creative architectural language. It was useful for developing presentation content and turning strategies into understandable design ideas.
Microsoft Copilot was useful for structure and simple practical summaries, especially in the original prompt stage.
The comparison shows that no single AI system should be accepted without review. The best result came from combining the strengths of each tool: Gemini for technical and contextual accuracy, ChatGPT for design communication, and Copilot for simple structured summaries. Final decisions still required human judgement to check whether the ideas were realistic, ethical, and suitable for climate-responsive architecture in Malaysia.