Architecture turns drawings and decisions into places that people may inhabit for decades. In this fifteenth essay in my 2024 research series, I consider how AI could influence that process and its visual culture. Generating an arresting building image is only the beginning; making a place work is a much wider task.
Site information, climate studies and knowledge of local history can inform a design. AI may help organise or explore that material, but data does not interpret a place on its own. Observation and conversations with the people who use it reveal priorities that a dataset may miss. A contextual design is more than a familiar local motif applied to a new façade.
Optimisation can compare alternatives against chosen goals, such as daylight or material use. The result depends on what is measured and what is left out. Curved forms are not inherently more intelligent or sustainable than simple ones. Structural analysis, environmental modelling and parametric design are also not automatically AI.
Tools may assist with documentation, specifications or preliminary checks, freeing time for design. Their output still needs verification against the actual project and applicable requirements. A plausible response does not establish compliance or safety. Architects, engineers and the wider project team retain responsibility for the decisions they issue.
Adaptable spaces could respond better to changing needs through layouts, controls or building services. That does not mean an AI system understands an occupant’s emotional life. Useful personalisation gives people clear choices, protects their privacy and allows manual control. In shared buildings, one person’s preferences must also coexist with others’ needs.
Material research could benefit from analysing test results and comparing alternatives. But predictions do not replace evidence about durability, fire performance, construction or maintenance. Environmental claims need a life-cycle perspective, including sourcing, replacement and eventual reuse. An unfamiliar texture is not proof of a better material.
Digital models and virtual walkthroughs can help clients, designers and engineers discuss a proposal before construction. These methods need not involve AI. Their value lies in making decisions understandable, including to people without specialist software. A polished simulation should reveal uncertainty as well as communicate an attractive vision.
After occupation, monitoring could help identify maintenance needs and support timely investigation. A prediction is not a guarantee that defects will be found. Inspection and professional judgement remain necessary. Repair and adaptive reuse may contribute more to a building’s future than repeatedly replacing it with a newer-looking alternative.
Cultural integration cannot be reduced to searching an image archive. Local craft, ways of living and contested histories require dialogue. Systems trained on widely circulated architecture may repeat an international visual vocabulary; deliberate choices are needed to leave room for less represented traditions and for genuinely local responses.
I see AI as a possible aid to architectural judgement, not a substitute for it. A richer visual culture would be measured not only by unusual forms, but by places that remain useful, understandable and worth caring for. The image of the future matters; so does the long life that begins after the image is built.

AI & Visual Culture: 16. Who Shapes the City? ↗

