AR / Anne-Catrin Schultz / Tanjšanje arhitekturnega prostora

Introduction: The Design Process and Changing Representations

The urgency to address issues related to climate change and other global challenges of the 21^st century is placing increasing pressure on architectural production to reconsider conventional approaches to both process and product. The need to conserve the planet’s resources and to work towards circular construction is accompanied by the availability of generative digital tools such as AI diffusion models that synthesize images from text-based prompts by composing pixels algorithmically. This essay traces the evolution of the role of drawings in design throughout history and examines the shifting conception of tectonics, linking both themes at the end to the shifts introduced by digital technologies such as AI image generation. [ 1 ]

While AI tools are emerging for all areas of life and production, AI-generated images might have a specific impact on architecture by moving finalized-looking images to the early schematic design phase, thus changing the format of articulating ideas and defining concepts. First ideas of buildings were traditionally developed through sketches, models, and conceptual diagrams. Architecture representation during the design process evolved from conceptual and abstract to more specific. Perspectives and renderings were typically produced towards the end of the process to share a “real view” of the project with clients and stakeholders.

AI-generated images are generated by deep learning models, typically trained on large-scale datasets comprising millions of images and their associated textual descriptions. When prompted to visualize an image of a speculative architectural concept, the model draws upon patterns learned during its training to synthesize new images. These generative models infer correlations between visual features and linguistic prompts, enabling them to produce outputs that reflect statistical associations between text and image. In response to a prompt related to architecture, the model may generate compositions that are visually compelling, surprising, and even inspiring. However, the results reflect a collage-like amalgamation of stylistic tropes and formal elements, echoing biases embedded in the training data. The statistical associations mentioned above are far from an architect examining precedents and exploring an iterative design process. The output is void of tectonic logic, material and structural performance, or construction methodology. This raises unresolved questions around ethics, intellectual property, and data governance. It might be necessary to adjust the training of designers and of AI models to develop competent new workflows for design ideation.[1] This essay traces the historical evolution of architectural drawing as a crucial part of the design process. It also examines the changing understanding of tectonics, linking both themes to the shifts introduced by digital technologies such as AI-driven image generation.

The conception of an architectural design ultimately leading to a building is a complex endeavor. Design is impacted by cultural forces, economic and functional requirements, regional conditions, and technological capabilities, all brought together by the architects’ and engineers’ approach. ‘Design’ as an activity represents a problem-solving act that is not limited to architecture but takes place in all professions. There is a specific goal, an appropriate solution, but the problem posed is multifaceted and must be defined and refined throughout the process. Building design frequently evolves from the abstract to the concrete, with precedents and previous examples leading the way. Depending on culture and timeframe, the drivers for architectural production (program, technology, expression, etc.) vary. In ancient Greece, for example, the form and configuration of the iconic Greek temples were given by a higher order. Spyro Kostof states: “The form of the temple must be God-given, and the recipient must be the highest representative on earth of divine authority. This often meant the king.”[2] Written instructions for buildings were archived on rolls of papyrus or leather to be consulted by the state architect as necessary. Rulers, squares, and triangles were used to develop and document the design. Drawings were drawn with reed pens on papyrus or leather.[3] Exploring further how design narratives and tectonic frameworks were generated and communicated throughout the architectural design process might illuminate how the increased resolution of formal expression is impacted by AI-image generation. Tectonics in this context is defined as the duality of technology, craftsmanship, and resulting form: the art and science of construction as an indissoluble unit.

Inspired by the architecture of ancient Greece, the Romans also valued proper proportions, believing that proportional relationships ensured both structural stability and formal harmony. Roman architects were less interested in the tectonic systems of their precedents –column orders were flattened as superficies and predominantly applied to facades, exemplified by the Colosseum in Rome.

Inspiration was largely drawn from built examples of Greek architecture and their reinterpretations, aligning politically and culturally with the cultural power communicated by them. The Roman architect, engineer, and theorist Vitruvius gives a sense of the range of drawings to be drawn during the design process: plans, elevations, and perspectives (which he refers to as scaenographia). Drawing tools—a compass and ruler—were used to set the outlines of the building. Elevations represented the façades and the building’s overall appearance. Beyond recording the geometric information, Vitruvius also refers to the term sciographia as “a perception of the building’s totality in depth, a view which reconciled the internal and external orders, the plan and the elevation.”[4] During Byzantium’s architecture production (the end of the Roman Empire), roles and responsibilities in construction started to diversify.[5] [ 2 ]

The Middle Ages, encompassing diverse phases and territories, from the decline of the Roman empire to the onset of the Renaissance—relied on a range of representational tools, including scaled parchment drawings, full-scale templates of details or building elements, and drawings at full scale, geometrically laid out with compasses and squares. Master masons and carpenters predominantly employed plan drawings to communicate geometric principles and to resolve constructive decisions and details. During this period, architects were less frequently mentioned than patrons—who were often credited with the realization of buildings—and skilled masons.[6] Architectural narratives, typically conveyed through ornamentation, were closely aligned with ecclesiastical authority or the governing aristocracy. The role of the architect evolved into that of the master-builder, a figure responsible for both the conception and supervision of construction. Knowledge and expertise were transmitted through guilds and trade organizations. Visual and tectonic expression emerged largely through the hands of craftspeople, often reflecting regional material practices and artisanal traditions. The rediscovery of perspective by Filippo Brunelleschi in the 15^th century—who is believed to have worked predominantly with wooden models—enabled a more effective presentation of spatial experience than was possible through orthographic drawing alone. Artists and architects took advantage of the tool, using section perspectives especially for building interiors. The manner in which the new tools impacted architectural design can be seen in Michelangelo’s “constructed” perspective of the Piazza Campidoglio in Rome [ 3 ] or Michelozzo’s design for the Tuscan city of Pienza. As with other tools and technologies, the use of perspective has significantly influenced the architectural design process by shaping the visual representation of space; however, it still necessitates the architect or designer’s control over formal expression and the skills needed to construct the drawings. In comparison to the architect’s and artist’s skills needed to master historic representational tools, AI image generation requires barely any training or expertise—machine learning takes on the generation of the images. This simplification has the potential to shift the architect’s role towards a mere curator of machine-generated content, losing agency and abandoning the iterative development of structure and form.

Historically, drawings became the primary medium of architectural communication, while perspective emerged as the essential technique for representing and designing architectural space—fundamentally transforming both design practice and architectural treatises. An impact of this magnitude can be expected from the mechanics of the AI-generated images as well, allowing for an easier amalgamation of different architectural languages and offering a set of new possibilities—and conundrums—for historicizing buildings. The design methods of architects changed with the onset of Humanism influenced by a push to return to antique ideals. Biographers confirm that the architect, goldsmith, and sculptor Filippo Brunelleschi traveled to Rome to learn from the principles of their architecture, working once more with antique precedents. It was common for architects to visit and study Roman ruins to fill sketchbooks with measured as-built drawings, such as the ones by Giuliano da Sangallo (1445−1515).[7] [ 4 ] These drawings were used to educate masons and builders back home, taking on the role of pattern books. Many others produced sketchbooks and drawing collections as well as theoretical writings to instruct architects about reviving antiquity. Treatises increasingly distinguished the architect developing buildings from the mason and carpenter executing them. Text and drawings operated simultaneously as instructions and a theoretical basis for architecture, while precedents served as examples to learn from. All these elements of the design process remain crucial when prompting AI to generate drawings or images. However, the elements are not directly emerging out of the designer’s personal study, experience, and evaluation. Technology has always been disruptive: the invention of the printing press fundamentally transformed the dissemination of information, making it accessible to a significantly broader audience. This technological breakthrough marked a revolution in information processing, comparable in scope and impact to the transformations currently unfolding with the rise of generative AI. Disegno, encompassing both drawing and perspective, emerged as the conceptual and practical foundation of the architect’s role, extending from ideation to the management of construction. As Alberto Pérez-Goméz notes, the “Renaissance architectural drawing was perceived as a symbolic intention to be fulfilled in the building, while remaining an autonomous realm of expression.”[8] The architect recognized “that the distance between idea and matter, between design and construction, would be reconciled through his own involvement in building.”[9] These observations highlight the enduring significance of precedents and representation—both visual and textual—throughout the history of architectural design.

Renaissance artists also cultivated the use of sketches as part of the design process. Leonardo da Vinci is credited with the trial-and-error mode of design,[10] using abstract sketches as the basis of the development of design and ideas. In the centuries following the Renaissance, architecture as a profession, discipline, and educational field became increasingly formalized. Architectural publications began to provide theoretical and methodological grounding for practice. The École des Beaux-Arts in Paris influenced not only architectural education in the 19^th century and beyond, but also shaped traditions of a detail-oriented, rendered version of architectural representation and thus deeply influenced the way ideas were transmuted, communicated, and visualized. Representations aimed at showing architecture in its monumental state and idealized reality. A curriculum based on competitions supported skill building and the advancement of visualization techniques. The selection of a ‘parti’ was crucial and would potentially guide the solution along a set of conditions given by the parti.[11] The Beaux-Arts representational conventions were replaced by a focus on technology and functionality as a driving force for design and expression. Architecture historian and theorist Dalibor Vesely writes:

“The primary conditions for a new relationship between architecture and technology were first established in the seventeenth century when a gap opened up between the traditional symbolic and instrumental representation. In this period, in the late seventeenth and the early eighteenth century, architectural thinking, which had always been closely associated through its long history with the mathematical representation of its principles, was overtaken by the new developments in the natural sciences.”[12]

This shift contributed to the establishment of engineering schools, which formalized the separation between structural and performance-based education and that of architectural design. Vesely identifies this period, beginning in the early 18^th century, as marked by the “growing arbitrariness of architectural decision making,” [13] a waning interest in classical traditions, and what he describes as a “discontinuity between the means and the content of representation.” [14] Despite these shifts, Vesely argues that the “dual nature of symbolic and instrumental representation was preserved in the cultural memory”[15] across centuries.

In the late 19^th and early 20^th centuries, various movements sought to reconcile technology with architecture. However, in doing so, technology itself often became the dominant force, displacing the layered architectural narratives of the past. The 20^th-century efforts to codify the principles of design into rational laws and rules have proven insufficient in capturing the qualitative dimensions that contribute to a building’s quality. Along with a strong belief in rational functionality and the precision of technical drawing, artistic movements, such as Cubism, introduced the concept of fragmental representation through collage in both fine arts and architecture. This approach allowed for the simultaneous presence of elements from different contexts and media. Pioneered by artists such as Pablo Picasso and Georges Braque, the tendencies around cubism included abstracted ideas of space by means of geometric shapes, frequently with a political leaning embedded. The technique of combining photography and drawing to convey conceptual and spatial ideas was adopted by Mies van der Rohe and many others, serving as a generative representational strategy across a range of architectural projects. Collage, as a design methodology, enabled the integration of diverse visual media and content into one composition. Overlapping possibly unrelated fragments of space are simultaneously perceived, a phenomenon later examined as literal and phenomenal transparency.[16] Le Corbusier and others would translate the layered concepts[17] into three-dimensional architectures, where overlapping spatial conditions could be experienced concurrently. While traditional graphic collages retain visible seams and discrete edges, AI-generated imagery tends to produce seamless integrations—collages in effect, but not in appearance. This shift, prefigured by the sophisticated use of digital tools like Adobe Photoshop, marks a transition in representational aesthetics from the visibly composite to the algorithmically synthesized.

The Reversal of Tectonic Syntax

Over the course of history, changes in formal expression have coincided with social evolution and technological progress. The early ideas regarding tectonics emerged from the analysis of Greek temples, where construction and expression were understood as inherently linked. The discovery that Greek temples were originally covered in a layer of paint (highlighting sculptural and architectural details) challenged prevailing perceptions of Greek architecture, revealing a layered system that told stories through both structure and ornamentation. These archaeological findings, along with the ongoing search for the appropriate architectural expression in industrializing societies, initiated the discourse of architectural tectonics. In the 19^th Century, German archaeologist Karl Friedrich Boetticher examined architectural expression based on underlying structural and functional necessity, differentiating ontology from representation. He argued that a building’s ornamental layer narrates its structural and material conditions, thereby uniting the two into a single entity. Decorative elements illustrate the mechanical forces or material conditions of the structure. Gottfried Semper expanded this discourse with a comprehensive examination of artistic language connected to practical and material conditions of craftsmanship. He traced the origins of architecture to textiles. For Semper, one of the most important elements of architecture is derived from the knot created to connect yarn or terminate woven fabric. Translated into built form, the knot became the joint or connection detail between components. The knot represented the beginning of all expressive details and a tool for material transformations that would reverberate through the forms of previous technologies, even when materiality had changed from soft textiles to brick and stone. In the 20^th and 21^st centuries, theorists continued the explorations around tectonics, expanding its scope from the building as a separate object to a larger context of global and regional conditions. [ 5 ]

The architect, theorist, and educator Kenneth Frampton is well-known for writings that examine tectonics at a time when industrialized construction dominated architecture production. Frampton clarifies that his study does not seek to deny the volumetric character of architectural form, but rather aims “to mediate and enrich the priority given to space by a reconsideration of the construction and structural modes by which, of necessity, it has to be achieved.”[18] He further emphasizes that the earthbound “nature of building is as tectonic and tactile in character as it is scenographic and visual, although none of these attributes deny its spatiality.”[19] This emphasis on the spatiality of architecture beyond its visual qualities is particularly noteworthy. Frampton initiated a renewed and more differentiated look at tectonics, re-introducing the poetic and symbolic notions of construction during a time when modern and postmodern architecture had revealed its deficiencies.

In 2006, architects Jesse Reiser and Nanako Umemoto published an ‘Atlas of Novel Tectonics,’ shifting the discourse from a static exploration to a dynamic one shaped by digital tools. They describe the transformation they observed: “We’ve gone from seeing temporal work in contrast to permanent architecture to seeing the temporal entering into the very fabric of architecture itself, rendering it ambient. […] We assert the primacy of material and formal specificity over myth and interpretation.”[20] For Reiser and Umemoto, the question no longer focuses on the question “What does this mean?” but “What does this do?”[21] While Gottfried Semper explored methods of making rooted in materiality and craftsmanship– along with the transformations of materials over time–computational design and digital fabrication methods have fundamentally altered how construction-related problems are approached and resolved. Therefore, the question becomes “how is this done?” Variables have shifted and continue to shift from specific products and trade conventions to parametric operations that are only recognizable as such but not readable to the user. Contemporary tectonic theories moved beyond the traditional expression of structural loads and material-appropriate craftsmanship to encompass the influence of digital tools and the environmental crisis, thereby linking the discourse to broader contemporary themes. Anne Beim articulates an ecology of tectonics, embedding “the concept of buildings as parts tied together as a whole in a broader context of natural and cultural systems.”[22] In the early 2000s, theories of “Digital Tectonics”[23] emerged, operating at “the intersection of the physical and the virtual.”[24] These theories explored the relationship between fabrication and the assembly of complex forms, merging tools of representation with those of production and construction. Branko Kolarevic notes, “The predictable relationships between design and representations are abandoned. The typological, curvilinear geometries are produced with the same ease as Euclidean geometries of planar shapes and cylindrical, spherical, or conical forms.”[25] Digitalization has also disrupted the predictable connections between structure, materiality and expression. In this context Rivka Oxman introduces the term “informed tectonics”[26] as a concept of material-based design, examining a shifting taxonomy and new digitally driven materiality. Antoine Picon describes the transformation in material culture resulting from digital tools as the ability to manipulate light and texture in an infinite combination of factors.[27] He observes that these surface conditions allow any image to be mapped onto facades and architectural elements, often devoid of the tectonic implications of physical materiality.[28] Picon also states that the gap between digital representation and traditional tectonics is not synonymous with a dematerialization of architecture, but merely a redefinition of materiality.[29] Picon advocates for a redefinition of design objectives and procedures, a new visual practice with the potential to navigate both local and global contexts. In times of crisis and instability the “distinction between abstraction and concreteness”[30] become increasingly blurred. This is further evidenced by a shift from abstract sketches and conceptual diagrams during the schematic design phase to fully rendered images of finalized buildings generated from brief textual descriptions. The evolution of tectonic theories mirrors the changing landscape of digitalization of the design process, which simultaneously reflects the cultural changes of everyday life. Again, it becomes evident that tools change process and outcome. As we remain immersed in the screens of our cell phones, our perception of materiality becomes almost constantly mediated by images. Architecture has not dematerialized but has absorbed some of the surface characteristics of the image-saturated world in which it operates.

Historically, the relationship between drawings and tectonics in the design process was closely tied to the moment of transition between representation to making, the handover to construction. In the industrialized content of the 20^th century, the design process typically followed a sequence of phases: schematic or conceptual design, design development, and construction or execution drawings. Perspectives and renderings were often produced later in the process, either to reassure clients that the project was progressing as intended or to market the building to its future audience. With the decline of trade conventions and traditional materiality—particularly the shift toward industrialized construction products—the generation of generative ideas (understood by the general public) became more arbitrary and individual. According to Mario Carpo, this first digital turn in the 1990s allowed the variability that architects, designers, craftsmen, and engineers craved. “Technology was meant to produce variations, not identical copies; customized, not standardized products.”[31] The road to mitigating the scale of standardized mass production and the embedded technical logic was and is long. The cost structure of digitally manufactured construction elements is different from the conventionally industrialized version. [ 6 ] [ 7 ] [ 8 ]

The Evolution of Artificial Intelligence in Architecture

The impact of digital tools on the architecture design process and ultimately on the built environment can hardly be overstated. Computers have transformed the visualization and production of architecture, ultimately changing the way architectural form itself is conceived and configured. Conceptual design in architecture has long served as a bridge between past experiences and contemporary problem-solving, enabling the development of new spatial and formal ideas. Precedent studies remain integral to architectural practice (and education), offering inspiration and insight from existing built works, while sharing symbolic content. Computer-Aided Design (CAD) revolutionized the design process and the production of drawings or models required for construction, eventually evolving into Building Information Modeling (BIM). The design sequence, once rooted in hand-drawn sketches on tracing paper and inked vellum, transitioned to digital representations of lines and volumes on screen. Artificial Intelligence also emerged in the 1950s and 60s as general problem-solving machines. Paving the road for the presence of AI in design and beyond, a report written at Stanford in 1962 by Douglas C. Engelbart[32] proposed that the integration of object-based design, parametric operations, and relational databases could elevate architectural problem-solving. While the report showcases the great potential of a general augmentation of the human intellect through the digital “clerk,” Engelbart uses an architect’s workflow as an example:

“Let us consider an ‘augmented’ architect at work. He sits at a working station that has a visual display screen some three feet on a side; this is his working surface and is controlled by a computer (his ‘clerk’) with which he can communicate by means of a small keyboard and various other devices. He is designing a building. He has already dreamed up several basic layouts and structural forms and is trying them out on the screen. The surveying data for the layout he is working on now have already been entered, and he has just coaxed the ‘clerk’ to show him a perspective view of the steep hillside building site with the roadway above, symbolic representations of the various trees that are to remain on the lot, and the service tie points for the different utilities. The view occupies the left two-thirds of the screen. With a ‘pointer,’ he indicates two points of interest, moves his left hand rapidly over the keyboard, and the distance and elevation between the points indicated appear on the right-hand third of the screen. […]”[33]

Engelbart’s vision of AI has, in part, materialized over the last decades. In 1969, Nicolas Negroponte voiced concerns that many might still share about architecture created by machines: “Computer-aided design cannot occur without machine intelligence — and would be dangerous without it. In our era, however, most people have serious misgivings about the feasibility and, more importantly, the desirability of attributing the actions of a machine to intelligent behavior. […] As soon as intelligence is ascribed to the artificial, some people believe that the artifact will become evil and strip us of our humanistic values. Or, like the great gazelle and the water buffalo, we will be placed on reserves to be pampered by a ruling class of automata.”[34] Negroponte also anticipated that the future fluid and capable machines would remove the barriers between architects and computing machines, changing the professional structures along the way. He writes: “With natural communication, the ‘this is what I want to do’ and ‘can you do it’ gap could be bridged. The design task would no longer be described to a ‘knobs and dials’ person to be executed in his secret vernacular. Instead, with simple negotiations, the job would be formulated and executed in the designer’s own idiom. As a result, a vibrant stream of ideas could be directly channeled from the designer to the machine and back.”[35] Negroponte’s vision appears strikingly accurate in light of today’s large language models. The 1990s saw the rise of neural networks and machine learning, which entered the mainstream discourse and application. AI has gained increasing visibility in architecture in the 2000s, with significant advancement and easier access in the 2020s. Neal Leach looks for the “logic informing these systems,”[36] which is necessary to comprehend the tectonic consequences of AI in the design process.

From Idea (image) to Building

The impact of the overwhelming presence of images has been the subject of commentary by theorists since the 1990s and early 2000s: Vilem Flusser observed the phenomenon of the “image flood,” (referring to photographs and film) foreshadowing the intensified concentration of images on our cell phones in the 21^st century. He writes, “We are accustomed, for example, to see the solar system as a geographic place in which individual bodies orbit around a larger one. We see it as such because it has been shown to us in images, not because we have perceived it with our own eyes.”[37] Architect and theorist Juhani Pallasmaa attributes the bombardment of visual imagery to mass consumerism and globalized economies.[38] He calls the profusion of images “a kind of suffocation in an endless Sargasso Sea of Images,”[39] giving rise to “an oppressive feeling of excess and eutrophication. He argues that the physical world, cities, and natural environments are all colonized by the (short-lived) image industry, questioning if the prevalence of the image pushes humankind back to prehistoric times of “gestures and images.”[40] Pallasmaa argues that the physical world has become a “pale reflection of the image,” possibly leading to the demise of imagination. For him, historic architecture was a vehicle to convey narratives and embody the meaning of stability far beyond individual consumption. As if foreseeing the intensified loss of tectonic intention through AI-generated imagery, he writes:

“Today’s forceful imaging techniques and instantaneous architectural imagery often seem to create a world of autonomous architectural fictions, which totally neglect the fundamental existential soil and objectives of the art of building. This is an alienated architectural world without gravity and materiality, hapticism, and compassion.”[41]

Pallasmaa refers to an absence of tectonics, a lack of material presence that is brought about by a reality increasingly dominated by images. Flusser’s image flood and Pallasmaa’s Sargasso Sea of Images have only expanded in scale. Prompting an AI image generation tool to suggest a design for a building initiates the process with an amalgam of references, seemingly collaged into a proposed visual. If the role of a conceptual drawing is the “embodiment of architectural ideas,”[42] what, precisely, is the role of the rendering or AI-generated image? Alberto Perez-Gomez asserts that: “Drawing is the architecture, a privileged vehicle for expressing architectural intentions: intentions that are poetic in a profound traditional sense, as poesis, as symbol making.”[43] The question of whether architectural design process can or should be based on AI-assisted image generation of final buildings evokes centuries-old debates about architectural production. Issues of design representation, construction technology, and tectonics remain deeply intertwined. AI-generated images are produced by models trained on extensive datasets of existing images paired with textual annotations, reflecting historical visual patterns rather than introducing novel creative concepts. Assuming a concept idea emerges through a written prompt, the simultaneous consideration of plan organization, access, sectional relationships, etc., must still to be addressed separately; they are not (yet) integrated into the tool.[44] The intentional message must come from the architect (or anyone else); political and cultural associations are difficult to trace unless explicitly embedded in the prompt. Architectural elements and forms drawn from a variety of contexts can’t convey a coherent cultural message – editing is necessary to maintain control of the narrative, especially since AI’s sources and inspirations remain completely unknown. AI-generated images represent a form of collage that is much more advanced than the fragmented collages of the early 20^th century design. The elements are fused into a seamless whole, and the layered overlaps that once allowed for the simultaneous existence of discrete objects and spatial narratives are no longer readily apparent.[45]

The rise of social media, the use of Pinterest boards (and other similar websites and applications) in design, and society’s increasing focus of society on imagery have contributed to the evaluation of architecture primarily through visual appearance rather than spatial experience. This shift is driven by the pervasive presence of images and screens, and perhaps by a lack of cultural and spatial education and public awareness. The dominance of the image and resulting loss of depth is not a new phenomenon and has been widely discussed. Alicia Imperiale, writing in 2000, identified the resulting reduction of depth and emphasis on surface as a profound issue. She uses the term “flat” to describe the paper or screen, and “surface” when referring to “issues that develop when architecture is built, when the emphasis shifts from the flatness of the representational space to the depth of the three-dimensional building.”[46] The ubiquity of images has appeared in façade design and architectural surfaces in a literal fashion, moving the attention from materiality to the surface. The façade as screen emerged in the early 2000s, with media facades and projections entering the architectural repertoire. The image appeared on facades as a thin layer printed on glass or the exterior skin. Facades as surfaces to print on, just as paper, also expressed society’s overall affinity to images. Herzog and De Meuron’s university libraries in Eberswalde and in Cottbus, both in Germany, illustrate the use of images on concrete and on glass. The skins of both libraries do not express structural forces or tectonic articulation; instead, they highlight the programmatic function of the buildings. The facades become commentary on the usage of the buildings, in both cases, their role as libraries and archives. The volumes follow a logic distinct from the visual language of the facades. In Cottbus, the building evokes associations of scrolls of paper. The architects refer to the glass skin imprinted with highly pixelated lettering as cladding the building “like a veil.”[47] [ 9 ] [ 10 ]

When employing an AI model that ingests broadly sourced, non-curated data from across the internet, the resulting image constitutes a synthesis of prior architectural forms and visual conventions. Rather than producing genuinely novel outputs, such models tend to recombine and recontextualize historical data, yielding visuals that reflect accumulated cultural artifacts more than intentional innovation. To move beyond mere recombination of existing patterns, the architect’s design literacy and creative agency become essential. A critical emerging skill for architects and designers will be the precise crafting of text prompts, paired with the ability to critically assess facades and models whose origins remain obscured. Political associations and consistent cultural messages will be left to the correcting hand of the designer, having to decode the conceptual framework instead of creating it. Conceptual principles that go beyond visual expression are difficult to embed. Gravity is not a factor in language-to-image generation, and the tectonic logic therefore has to be applied after the fact, which is the opposite of a process that starts with an abstract geometric, functional, and contextual concept that expands in specificity through alterations.

Concluding Remarks

In this examination of the potential impact of AI-generated images on the architectural design process and its outcome, it becomes evident that architectural design has historically relied on the interplay of images and precedents to reinvent expressions, typologies, and spaces. As building conventions shifted across societies and time frames, they involved the re-crafting of embedded messages into new narratives, alongside evolving technologies of representation and construction. A similar impact is to be expected from digital tools, particularly as they transition from replicating pre-digital tasks (drafting conventional sets of drawings for design communication, engineering coordination, and construction) to actively shaping design processes and tectonic possibilities. The consequences of a digitally driven design process have already led to profound changes in architectural expression. In some cases, this has resulted in dynamic parametric designs; in others, it has contributed to a thinning of architectural space and substance, raising questions about the tectonic presence of materials and structures. Digitalization has enabled design approaches that rely on algorithms and parametric modeling, establishing relationships between elements through a range of parameters. Yet, the spatial understanding of the 21^st century continues to be informed by modernist ideals and the legacy of industrialized construction methods.[48] With the image flood discussed above, now amplified by vast quantities of AI-generated representations, considerations around architectural space risks becoming obsolete. Architects may unintentionally relinquish their authorship of the architectural narrative that has historically allowed them to embed both symbols and critical commentary related to politics and social conventions.

Architecture increasingly becomes image, potentially communicated through what might be termed machine hallucinations. Space hasn’t been explored as a resource for contemporary society with the same dedication as visual representation in contemporary discourse. Michael Hensel observes, that while space (naturally) continues to play an active role, “our repertoire of spatial concepts and our ability to understand and work with them remain relatively underdeveloped compared to the formal innovations, programmatic savvy and critical sophistications of the past decades.”[49] What Douglas Engelbert referred to the “augmented architect” demands a transformation in education and skillsets, one that reconciles the internal and external orders of architecture. While future AI software may significantly expand architecture’s generative potential—including spatial and organizational configurations, code compliance, and engineering coordination–the uncritical use of finished imagery derived from unknown and uncited sources poses an opportunity for novel expression and risk of losing all intentional narrative. The lack of focus on spatial qualities might get intensified by AI-generated solutions, or the confluence of all digital tools might lead to profound innovation and a renewed spatiality. Counteracting a tempting invitation to forego the critical evaluation of the evolving tectonic conditions of architecture, all emerging technology must be assessed for its capacity to support the built environment in expressing human culture and addressing the challenges of the 21^st century.