From Rendering to Semantization: Intensified Architectural Semantics with Generative AI

Abstract

Generative AI is often framed in architecture as a faster rendering engine, able to turn sketches, models, and prompts into persuasive images. uch a view collapses two distinct operations: rende­ring and semantization. Rendering concerns ap­pearance; semantization concerns the process through which geometry becomes meaningful, legible, and recognizable as architecture. Building on this distinction, the article proposes three re­gimes of semantization: natural, anticipatory, and intensified. While natural semantization emerges through use, material identity, and time, and anti­cipatory semantization is associated with architec­tural rendering, intensified semantization names the condition introduced by GenAI, in which me­aning is accelerated, thickened, and partially de­stabilized through probabilistic image formation. Methodologically, the paper develops sequen­tial protocols through two case studies. The first starts from a deformed carton package and follows its transformation from everyday object to neutralized meta-object, then to anticipatory ar­chitectural rendering, and finally to three bran­ches of intensified semantization: typological, atmospheric, and affective. The second begins with the Barcelona Pavilion and subjects it to con­trolled semantic displacement in order to test th­resholds of recognizability and defamiliarization. Across both cases, the article introduces semantic inflation as the failure mode in which intensifica­tion becomes overcoded, literal, or cliché-driven. The paper concludes that GenAI should not be understood only as an instrument of improved visualization, but as a semantic laboratory for architecture. It enables branching, iterative, and comparable transformations through which ar­chitectural meaning can be produced, displaced, and evaluated.

DOI: https://doi.org/10.20365/disegnarecon.36.2026.9