Freehand drawing in the first year of architectural education is often shaped by a fundamental gap between seeing and drawing. Novice students encounter space, volume, line, depth, texture, light, and material qualities, yet they do not always know how to translate these visual experiences into architectural drawing, sketching, rendering, and representational judgment. This study addresses this pedagogical gap by examining the use of artificial intelligence in teaching freehand drawing to undergraduate architecture students. Rather than approaching AI as a substitute for manual drawing or as a tool for producing final visual outputs, the study conceptualizes it as a visual mediator that can make the implicit logic of drawing more visible, discussable, and pedagogically actionable. The main aim of the research is to formulate and evaluate the “AI-based Visual Mediation Model” as an instructional approach for strengthening students’ representational understanding in architectural freehand drawing.
The research was conducted through a design-based educational research approach, supported by qualitative and descriptive analysis, in the Architectural Representation I and II courses at Shahid Beheshti University during the 2025–2026 academic year. In the first semester, one group of 16 students out of 62 first-year architecture students participated in the initial phase of the intervention. In the second semester, three groups, comprising approximately 37 students, were included in the observation and instructional intervention process. The educational intervention was implemented across selected sessions within the two-semester course sequence and involved a gradual movement from direct observation and manual drawing toward AI-generated intermediate representations, guided analysis, redrawing, rendering, and final visual organization.
The data included students’ initial and intermediate drawings, AI-generated visual mediations, the instructor’s observational notes, classroom feedback, analytical rubrics for freehand drawing and rendering, and expert evaluations of selected final works. The findings indicate that, when accompanied by the instructor’s analytical guidance, AI-generated intermediary images can help students recognize the structural logic of lines, vanishing directions, spatial depth, foreground-background differentiation, texture control, preservation of white space, tonal hierarchy, and purposeful rendering. The model was particularly effective in shifting students’ attention from merely copying visual appearances toward understanding the selective and interpretive nature of architectural representation. It also supported the transition from isolated drawing exercises to more complex visual compositions, including rendered sheets and organized architectural presentations.
The study concludes that the pedagogical value of AI in this context does not lie in accelerating image production or replacing the hand, but in altering the quality of students’ visual attention. In the proposed model, AI functions as an intermediate visual layer through which students can compare, question, analyze, and return more consciously to manual drawing. Accordingly, AI becomes not an alternative to freehand drawing, but a means of deepening the relationship between observation, interpretation, and hand-based architectural representation.