I learned that AI output is highly dependent on how prompts are written. Clear and structured language produces more accurate and relevant results, while vague or overly visual prompts can lead to unexpected interpretations.

I also learned that different AI tools serve different roles in the design process. Text AI is stronger in conceptual development, while image AI focuses on visual interpretation, and research AI provides contextual knowledge.

In the text AI tool, I observed that the system can sometimes shift between different output types, such as text and image generation, depending on how the prompt is structured. This showed that AI does not always strictly follow the intended format unless clearly instructed.

In the image AI tool, I noticed inconsistencies in spatial interpretation. For example, when I intended a single vertical tower, the AI initially generated two separate towers, showing that it can misread architectural unity and composition.

These behaviours show that AI interpretation is not fixed and can vary significantly between tools and prompt styles.

AI lacks true spatial understanding and architectural reasoning. It often focuses on visual output or pattern recognition rather than functional or contextual design logic.

In both text and image tools, AI can misinterpret architectural intentions, especially when dealing with complex spatial systems like vertical circulation or layered programs.

This highlights that AI should be used as a supporting design tool rather than an independent designer, as it still requires human control, interpretation, and critical evaluation.