Over the last few months we’ve started to notice small shifts in how some clients approach architecture. The signals are still faint, but they’re getting more frequent, and they point to a change moving fast as AI lands in the sector.
Some of it shows up as unexpected conversations. Clients question the building code with a confidence they didn’t have before, backed by arguments that weren’t part of the usual dialogue. “If the corridor only leads to a bathroom, it can be 60 cm wide,” they tell us. The line sounds convincing in the abstract. Regulation and construction don’t actually work that way.
The rest shows up in the inbox. Briefs now arrive with full AI-generated dossiers attached: floor plans, electrical schematics, plumbing proposals, even material specs. Well presented, coherent at first glance, and full of basic errors once you read them properly. Layouts that don’t function, dimensions that breach the building code, solutions that ignore how a built space behaves.
What both cases have in common isn’t the content. It’s the source. Answers pulled from AI systems that present themselves with total certainty, as if fluency were the same thing as accuracy.
Who hasn’t used an AI tool by now? Most of us have asked ChatGPT or Gemini to settle a quick doubt. These tools earn their place: they help you grasp a concept, organise information, and test ideas on the spot.
That same speed has a side effect, though. When an automatic answer gets taken as valid without any check, it can warp a decision from the very start of a project. It hits hardest on questions of regulation, planning, and feasibility.
The danger of AI in architecture isn’t the tool. It’s how people read its actual reach.
Understanding AI’s real impact on architectural practice
Consultation tool vs. technical judgment
AI works as a starting point. It doesn’t replace the technical judgment a specific case demands. In architecture almost no answer is universal: it depends on the urban context, the planning rules in force, and how those rules get read.
We see this constantly. Even the conversations with the district technicians who review projects rest on interpretation of the regulation. The personal element carries real weight in how technical criteria get applied.
Planning regulation isn’t a closed system. It’s an interpretive framework, and it needs experience, technical cross-checking, and actual sign-off from the administration.
The RIBA AI Report 2025 backs this up. Adoption among UK architecture practices jumped from 41% in 2024 to 59% in 2025, but RIBA frames AI’s role as augmenting professional work rather than replacing it, and flags real concerns around control, accountability, and technical decision-making (RIBA AI Report 2025).
AI, medicine, and architecture: comparable limits
The limits get clearer next to other complex fields. AI won’t replace a full medical diagnosis or a clinical assessment, and it won’t settle an architectural or planning question for good either. It can sketch a first outline. It can’t stand in for a complete analysis of the case.
In architecture that complexity comes from layers you only understand from practice:
- the specific regulation of the municipality and how it’s actually interpreted
- experience built up across similar projects
- direct observation of the space and its physical constraints
- reading the urban surroundings and how they behave
- dialogue with municipal technicians and the administration
- unwritten criteria you only learn by doing the work
Strip those out and any answer is partial, with a real risk of misjudging whether a project is even viable.
The illusion of instant knowledge
AI produces fast, tidy, confident-looking answers, and that polish can hide how involved the architectural process really is. The cost lands in the early stages, where you set expectations that are hard to walk back later.
General information isn’t the same as the regulation that applies to your case. A digital example isn’t proof your plot is viable. A render isn’t an approvable project.
RIBA names this same tension. AI is being adopted fast for analysis, drafting, and design support, and at the same time it introduces errors from missing context and a tendency to over-rely on automated output. Hence the case for constant human oversight.
Risks in interpreting planning regulations
Regulation as an interpretive system
One of the most common mistakes is treating planning regulation like a closed manual. In practice, a large part of the technical work is reading how that regulation applies to each specific case.
That reading often isn’t written down anywhere. It gets built through experience, through earlier projects, and through direct contact with municipal technicians.
Using AI uncritically for regulation and feasibility
AI tools have no access to local criteria, administrative precedent, or the practical interpretations of each municipality. So they can hand you answers that look right and aren’t, once you put them in a real context. An AI answer doesn’t replace a planning consultation or a technical sign-off on the project.
RIBA makes a related point: practices need clear policies on how they use AI, precisely to avoid these interpretation errors and keep technical accountability intact through the design process.
Theory vs. administrative reality
Regulation can read as perfectly clear in the abstract. Applying it depends on the urban context, the type of intervention, and how the administration interprets it. That’s why two similar projects can resolve in completely different ways.
Practical problems in decision-making
Confusing information with validation
A frequent error is assuming an AI answer equals an official confirmation. That gap creates expectations the project’s technical and legal reality won’t support.
Underestimating the complexity of the architectural process
A project isn’t a visual or a concept. It runs through technical, regulatory, and administrative stages that decide whether it’s viable at all: the planning analysis, the technical development of the project, the consultations with the administration, and the adjustments that come up during licensing.
Unrealistic project expectations
AI-generated images and explanations can be genuinely persuasive. They aren’t always compatible with the regulation or with the real construction limits of the site.
The role of technical judgment in a project
Interpretation as an essential part of the process
The value of the architectural process lives in interpreting the context and the regulation, not in reading either one literally. That interpretation is what turns an idea into a buildable project.
Validation with technicians and the administration
Plenty of decisions only get confirmed through accumulated experience, consultation with municipal technicians, and real work on comparable projects. Automated systems don’t hold that knowledge.
From quick information to responsible decisions
Architecture turns information into decisions you can stand behind, balancing expectations, regulation, and what you can actually build. What matters isn’t what AI suggests is possible. It’s what’s viable inside the technical, legal, and urban framework.
Strategies for a clearer relationship with your project
Use AI as a reference, not validation
Lean on AI for support. Don’t let it stand in for technical verification or for an informed decision.
Understand the invisible phases of a project
Behind every project sits technical work you rarely see: the regulatory analysis, the consultations, the adjustments, the progressive sign-offs.
Accept complexity as part of the process
Architecture is complex by nature. Working with that complexity, rather than around it, is what gets you to decisions that are realistic and safe.
FAQs
Can you trust AI to understand planning regulation?
Only as general orientation. It doesn’t replace the technical interpretation or the sign-off your specific case needs.
Why doesn’t the information online match what you tell me during the project?
Because regulation gets applied in context. It depends on technical criteria, experience, and how the administration handles it in practice.
Is regulation always fixed?
No. It’s interpretive, and it shifts with the municipality, the urban context, and how it’s technically applied case by case.
