Testing ChatGPT-5's Critical Ability on Spatial Questions

Today I tested ChatGPT-5 in the field of topological criticism by asking it to prepare a 5-minute academic critical review of my 2014 paper From Space to Place: A Necessary Paradigm Shift in Architecture, where I first outlined the hypothesis of place understood as a systemic and processual notion (place as system of processes).

I then transformed the resulting text into a podcast format, with the aim of presenting these ideas in a concise and accessible manner. This experiment highlights the potential of AI not only as a tool for analysis, but also as a medium for communicating complex architectural and philosophical concepts to a broader audience, provided that its output is carefully mediated and guided by human judgment.^[1]

Podcast: ChatGPT-5’s Critical Review of From Space to Place: A Necessary Paradigm Shift in Architecture.

Note

[1] The first attempt made by ChatGPT-5 was rather unsatisfactory, not to say disappointing: it completely missed the point of my original paper, filling the review with commonplaces that neither belonged to my original conception nor reflected my vocabulary, seemingly drawing arguments from elsewhere on the web rather than engaging with my text. I rated it a ‘very bad review.’ Without any additional prompting on my side, as an answer to my rating, the AI then independently revised the script in order ‘to better engage’ with the paper, this time producing a version that ‘pointed out both strengths and weaknesses with more accuracy.’ This second attempt—the one featured in the podcast—is much more compelling and, overall, quite faithful to the original text.

It is not my intention here to engage in a point-by-point debugging of ChatGPT-5’s criticism. Overall, the entire RSaP—Rethinking Space and Place project can be viewed as an extended answer to ChatGPT-5’s legitimate criticisms. However, I found its review quite human in substance. In particular, points one (over-idealization of place), two (risk of vagueness), and four (implementation/feasibility gap), although essentially addressing the same underlying issue—the practical implementation of the theory—closely resemble critiques that fellow architects have raised about my work, both in the past and more recently. As a matter of fact, it is worth stressing that the central proposition of the theory expressed in that paper—place as a system of processes—emerged directly from practice, not from abstract speculation (see my Badel Block Redevelopment Project, Zagreb, which I also mentioned in the paper)—a question that the AI still overlooked. Moreover, the paper was inevitably constrained by the academic context in which it was presented. As I wrote in the article introducing this website, Rethinking Space and Place, this platform was created precisely to continue exploring the questions raised in that early paper but which I could not then develop in detail. For architects particularly interested in the question of implementation of the theorethical framework behind the notion of place understood as a systemic and processual notion, I refer you to The Place of Architecture The Architecture of Place – Part III: A Case Study.

Finally, ChatGPT-5 asked me: ‘Would you like me to also draft a closing paragraph that frames this AI interaction as a broader reflection on the role of artificial intelligence in architectural theory and criticism?’ Yes…

AI: ‘This episode also illustrates a broader point: artificial intelligence, while often prone to generalizations and simplifications, can serve as an unexpected interlocutor in theoretical debates. Its ability to reproduce, reformulate, and even anticipate critiques already voiced within the discipline demonstrates both the limits and the potential of AI as a tool for reflection. For architectural theory, where questions of abstraction, practice, and implementation are always intertwined, such exchanges may open new forms of dialogue—forcing us to refine our arguments, clarify our concepts, and test their resilience against different interpretative frameworks. In this sense, the critical encounter with ChatGPT-5 becomes not merely an experiment in automated reviewing but a reminder of the evolving conditions under which architectural and philosophical thought unfolds today.’