Clients are using AI image generation tools (primarily Midjourney, DALL-E, and ChatGPT) to modify architectural renders and send the results back as design feedback. This is a shift in how clients communicate preferences, not a threat to the architect's role. The architect's job is to translate that input into buildable architecture, which is exactly what AI cannot do.

By the numbers

• Architecture firms using AI jumped from 41% in 2024 to 59% in 2025, an 18 percentage point increase in a single year — Royal Institute of British Architects (RIBA), RIBA AI Report 2025, June 2025.
• Only 8% of architecture firms have implemented AI solutions into their practice, while 84% of architects believe AI can automate manual tasks to save time — American Institute of Architects (AIA), Artificial Intelligence Adoption in Architecture Firms: Opportunities and Risks, 2025.
• 77% of Design and Make industry leaders plan to continue increasing investment in AI and emerging technology over the next three years — Autodesk, 2024 State of Design and Make Report, 2024.
• Construction labor productivity has grown just 1% annually over the past 20 years, compared with 2.8% for the total world economy — McKinsey Global Institute, Reinventing Construction Through a Productivity Revolution, 2017.

I was on a demo call last week with an architecture firm in Jordan. Mid-conversation, the principal told me something I wasn't expecting.

"We started getting feedback from clients," he said. "But it wasn't normal feedback. They were taking our renders, the ones we'd sent them, and running them through ChatGPT. Asking it to change the entrance. Modify the bedroom layout. Create a different feel for the space. Then they'd send that back to us as direction."

He paused. "At first, I didn't know what to do with it. Who's the designer now? What is my role? Really, what should be our stance on this issue?"

This isn't a theoretical future scenario. It's happening right now.

What is the new client collaboration pattern with AI?

Younger clients (he mentioned both were in their early 30s) are treating AI tools the way previous generations treated Pinterest boards or magazine tearsheets. They're not trying to replace the architect. They're trying to communicate what they want in a language that feels natural to them.

The architect's instinct was to feel threatened. Wrong frame.

When clients used to bring in magazine cutouts or Pinterest boards, nobody saw that as an attack on design authority. It was input. A starting point for conversation. A way to understand aesthetic preferences before translating them into buildable architecture.

AI-modified renders are the same thing, just faster to produce and more specific in their edits.

What does this mean for architects?

The real question isn't whether clients will use AI in architecture. They already are. The question is how you maintain design authority while incorporating AI-modified client input into your process.

Rejecting it outright doesn't work. Saying "that's not how we work" or "AI doesn't understand architecture" just creates friction. The client has already invested time exploring their preferences. Dismissing that feels dismissive of them.

What does work: treating it as a design conversation starter.

"I see you've explored moving the entrance. Let's talk about why that appeals to you, and what that would mean for circulation, structural logic, and the rest of the design."

You're not accepting their AI output as the final design. You're using it as a window into their preferences, then bringing your expertise to evaluate feasibility, code compliance, constructability, and design intent.

What does the client actually think when they modify your renders?

Put yourself in the client's position. They're not architects. They can't sketch. They can't model in 3D. They struggle to articulate spatial preferences in words.

For decades, the best they could do was point at references and say "something like this." Now they can generate variations quickly and show you exactly what they're imagining. For the first time, they can be specific.

From their perspective, they're being helpful. They're doing the work of exploration so you don't have to guess.

The problem comes when architects read this as the client trying to do the architect's job. That's not what's happening. They're trying to communicate more clearly.

How architects should handle AI-modified client feedback

1. Acknowledge it without defensiveness.

Don't say: "Well, AI doesn't understand structure, so that won't work."

Say: "Interesting, you're drawn to moving the entrance to the east side. Let me understand what appeals to you about that, and then I'll show you what that enables (and what trade-offs it creates)."

2. Use it as a design thinking tool.

Treat the AI-modified render as a prompt. What is the client responding to? Color? Layout? Proportion? Spatial relationship?

Often the client isn't actually attached to the AI output itself. They just couldn't articulate what they wanted before they saw it. Anyone who's sat through a client meeting where the feedback is "I don't know, it just doesn't feel right" knows how valuable that kind of specificity is.

3. Bring the architect's expertise into the conversation.

"You've moved the entrance, which is interesting. That changes the circulation sequence. Here's what happens to the flow through the rest of the house. Here's how it impacts structural logic. Here's what it does to natural light. Let me show you two options: one that captures what you're after while keeping the design coherent, and one that explores a different approach to the same goal."

You're not saying no. You're doing the job the AI can't: connecting their preference to what the building can actually do.

4. Document your reasoning.

Clients using AI are often more engaged than clients who passively approve whatever you show them. They're thinking about the design. That's good. But it also means you need to be clear about why certain things work and others don't.

This isn't about defending your design ego. It's about educating the client so they understand the decisions that go into making a building work.

How should architects respond to AI-modified client feedback?

Architects using Snaptrude, an AI-powered, cloud-native BIM design tool, can generate multiple design options quickly, giving clients more to react to and less reason to modify renders on their own.

Try Snaptrude free →

What deeper shift is AI creating in the architect-client relationship?

For most of architecture's history, clients were passive. They'd see Option A, Option B, Option C and pick one. The architect controlled what was even possible to imagine.

AI is shifting that dynamic. Clients now have access to tools that let them explore on their own. They can generate variations. They can show you what they mean instead of trying to describe it.

That doesn't diminish the architect's role. It changes it.

You're no longer the only person in the room who can visualize spatial ideas. But you're still the only person who understands structure, code, constructability, material performance, cost implications, and how all those constraints interact with design intent. That part hasn't changed, and it won't.

AI lets clients communicate more clearly. Architects translate that into buildings that actually work.

What should architects actually do about this?

Your clients are already using AI on your work. The question isn't whether to accept that reality. It's how to use it as input into better design conversations.

Here's how to get ahead of it:

1. Build an internal framework for handling AI-modified feedback. Don't let every project team figure this out independently. Create a shared approach: how to respond, how to frame the conversation, what language to use.

2. Train your team not to be defensive. The instinct is to protect design authority by rejecting AI input. That's the wrong move. The right move is to welcome client engagement and then show why architectural expertise still matters. There's a difference between letting someone explore and ceding control.

3. Use it as an opportunity to differentiate. Most architects will resist this. If you're known as the firm that welcomes client exploration and then makes it buildable, you'll stand out. Tools like Snaptrude, an AI-powered, cloud-native BIM design tool, let your team generate and iterate on design options quickly, so you always come to client conversations with more to show.

The architects who figure out how to work with client exploration, rather than fight it, will build stronger relationships and win more work. The ones who push back will spend their time arguing about process instead of designing buildings.

Try Snaptrude free →

Frequently Asked Questions

Should I tell clients not to use AI on my work?

You can try, but it's neither enforceable nor productive. The better strategy is to treat AI-modified feedback as client communication, not a challenge to your authority. Clients who take the time to explore their preferences visually are more invested in the outcome than passive clients. The goal is to use that input to have a better design conversation, not to shut it down.

What if the AI-modified design violates code or structural logic?

That's your clearest opportunity to show architectural value. Walk the client through why the AI-generated move creates a structural, code, or constructability problem, then propose two alternatives that achieve the same spatial or aesthetic goal without those issues. You're not shutting down the idea. You're translating it into something buildable, which is precisely what clients hire architects to do.

Does this mean architects will be replaced by AI?

No. A client generating AI render variations can't replace an architect's ability to integrate structure, MEP systems, code compliance, cost constraints, constructability, and design intent into a building that can actually be permitted and built. The gap between an AI-generated image and a coordinated set of construction documents is enormous, and every step of closing that gap requires architectural judgment. AI generates images. Architects design buildings.

How do I maintain creative control if clients are exploring on their own?

You maintain creative control by being the person who evaluates feasibility and translates visual preferences into buildable architecture. A client can generate 100 AI variations, but only you know which spatial moves are structurally viable, code-compliant, and constructable within the budget. Creative authority isn't about controlling what clients imagine. It's about having the expertise to make the best version of what they're imagining.

What tools are clients using for this?

Primarily image generation tools: Midjourney, DALL-E, and ChatGPT's image generation features. Clients typically upload a rendered image and prompt changes to color, material, massing, or spatial arrangement. These tools aren't architecture-specific, which means their outputs routinely ignore structural logic, MEP constraints, and code requirements. How useful the output is as a design communication tool depends on the architectural sophistication of the prompt.

How are architecture firms changing their design workflow in response to clients using AI?

Firms adapting most successfully are increasing the frequency of early-stage design sharing rather than presenting polished options only at milestone meetings. Showing more options earlier, even rough ones, reduces the window where clients feel compelled to explore on their own. Tools that make generating multiple design options fast, such as cloud-native BIM platforms, make this higher-frequency sharing workflow practical without consuming additional drafting time.

How does Snaptrude help architects manage AI-modified client input?

Snaptrude, an AI-powered, cloud-native BIM design tool, lets your team generate multiple design options quickly inside a shared, cloud-based model. When clients can explore variations with you in real time, they have less reason to modify renders on their own. It keeps the design conversation in the right hands while still giving clients the visual responsiveness they want.

‍Try Snaptrude free →