Most Architects Now Use AI. Here's What They're Actually Doing With It.

AI isn’t an experimental future headline anymore, it’s infrastructural. It’s in our browser tabs, it drafts emails and summarizes meetings, recommends what to watch or buy and how to price risk. It’s apparently everywhere. 

Construction, however, doesn’t run on algorithms. It runs on liability, regulation, coordination and an undisputed physical reality. A rendering can impress a client. A structural miscalculation can cost millions. So the appropriate question would be: how deeply has it entered architectural practice?

We have a clear answer across ten European countries. As of Q4 2025, 59% of architects report using AI tools in their design work. That’s a compelling majority. Despite the day-to-day workflows for more than half of European architects – the data shows adoption is uneven, use cases are concentrated, and enthusiasm is filtered through professional judgement.

Let’s go through the specifics.

Adoption Is Wide — But Far From Uniform

The overall AI usage rate among architects stands at 59% for both regular and occasional use. Underneath that average, there’s significant variation:

• Denmark – 76%
• Spain – 70%
• Poland – 64%
• UK – 55%
• France – 36%

Let’s look at the polarity – Denmark and France. Same profession, same continent, forty percentage points apart.

In Denmark, AI is clearly routine in many offices. In France, a sizeable share of architects still hasn’t brought it into their workflow. That gap isn’t theoretical – it changes how quickly habits shift, how firms organize work and how clients are served. Countries with strong BIM penetration and digitally confident firms are moving faster, more fragmented markets are moving cautiously.

Age and company size play a role too. Architects aged 18-34 are much more likely to use AI than older counterparts. Larger firms adopt more readily than small studios. Bigger teams tend to test tools earlier, partly because the productivity upside scales faster.

So yes, AI is mainstream in architecture. But it’s far from uniform.

What “Using AI” Actually Means

When architects say they use AI, what are they referring to? 71% of them tell us in no uncertain terms – ChatGPT is the common denominator. A country-by-country breakdown shows some variation in usage points, but the lowest values are still over 50%.

Other tools show up, but they’re secondary:

• Gemini – 18% overall
• Copilot – 9% overall
• Midjourney – 2% overall, with far stronger numbers in Spain and Denmark
• Firefly – 2% overall, but best represented in Italy

The pattern is clear: AI adoption in architecture is not being driven by specialized design automation software. It’s being driven by flexible, browser-based generative tools that are easy to access and easy to test. That keeps experimentation high but it also means deep integration into BIM workflows remains limited.

Where AI Is Used – and Where It Definitely Isn’t

The strongest use cases are practical and contained.

49% overall use AI to generate technical documentation. It’s repetitive, structured and often time-consuming. So when a tool comes along that helps draft, structure and refine text – we welcome the lightened load. The architect still reviews everything, responsibility stays put. But the first draft comes faster. That’s a no-brainer for efficiency.

We’ll see renderings and visualization in a close second at 43%, with Denmark and France – again – being the polarities. Visual output is where AI makes an immediate difference, because alternatives can be generated quickly, early ideas can be visualized without long production cycles and client discussions move way faster. Visibility saves time.

Early concept and massing support is growing, particularly in the more digitally confident markets. Still, it’s not dominant.

The technical end of the workflow is a different tale altogether. Automating repetitive modelling tasks, creating automatic BIM objects, technical or quantitative data analysis… single digits, barely. That tells us AI helps around the edges. It can’t rewrite the central process. 

Productivity Gains: Real, Not Radical

Architects were also asked about time savings.

Modelling

• 17% report significant time savings
• 47% report small time savings

That leaves a substantial share seeing limited modelling impact. In the Netherlands, 38% report significant modelling time savings. Belgium, Spain, Italy, Germany and Denmark cascade in percentage points. So, helpful gains? Yes. Complete shift in how modelling works – of course not.

Documentation

• 30% report significant time savings
• 33% report small time savings

Why 41% Opt Out

If 59% are using AI, 41% aren’t. The main reason is that 44% of them see no clear benefit to it.

• Belgium – 60%
• Denmark – 55%
• France – 35%
• Germany – 16%

Other reasons include lack of knowledge or time to learn, satisfaction with existing tools and reliability concerns. The key point is: architects are not rejecting AI, but assessing it. If it doesn’t clearly improve project delivery, it doesn’t get priority. That’s how a liability-driven profession behaves.

Where Architects Sit in the Wider Market

Compared to other construction professionals, architects are clearly ahead of the race in AI usage. Contractors trail behind. Installers and painters report minimal uptake. This state lines up with where AI currently delivers value. Ideation, visualization and documentation are upstream activities. On-site execution, coordination and installation depend more heavily on precision, sequencing and physical constraints.

AI’s strongest foothold today is at the front end of the value chain.

Conclusion

AI is now a working tool in European architecture. A majority of firms use it. In some countries, it’s deeply embedded, in others – it’s still under evaluation. Architects are applying AI where it improves output and ignoring it where it doesn’t justify the risk or the effort. That’s not caution for its own sake, it’s more akin to professional accountability.

So here’s the lowdown for Q4: AI is established, but not transformative. It’s a productivity layer, not a structural overhaul. Adoption is strong, uneven and selective. And unless the technology proves clear value inside technical production workflows, that balance is unlikely to shift dramatically in the near term.

The direction is clear. The pace is controlled.