The IPQ Field Guide
AI search for architects
A client with a project and a budget now types the question into ChatGPT before they type anything into Google. So we ran the experiment nobody in this industry has published: we asked four AI engines the four questions that come before hiring an architect, and logged every single source each one used to build its answer.
275 cited sources across 97 domains, pulled on 10 September 2026. Awards did not show up and neither did the design press. What the machines read, over and over, was other architecture firms' websites, and always the same kind of page.
The short answer
Across 16 answers from four AI engines, 63% of the 275 cited sources were a practice's own website. Directories and marketplaces together accounted for 10%, and professional associations 9%. Every practice page the engines cited was a guide answering the client's question, with titles like "how to choose the right architect for your home" and "how to choose a commercial architecture firm for your project". Not one portfolio page, project gallery or services page was cited anywhere in the set.
What we measured, and how
Four questions, split evenly between the two buyers. Two residential: how to choose a good architect for a custom home build, and what an architect costs for a renovation or addition. Two commercial: how to choose an architecture firm for a commercial building project, and what a developer should look for when selecting an architect for a new development.
Each question went to four engines with live web search enabled: Perplexity (sonar), ChatGPT (gpt-4o), Gemini (2.5 Flash) and Claude (Sonnet 4.6), through the DataForSEO AI Optimization API on 10 September 2026.
We kept every citation each answer returned, resolved it to a domain, and classified the domain by who owns it. We then pulled seven live US Google results for the same buyer questions to see what Google's own AI Overview was citing.
63% of what the AI reads is another firm's website
| Source type | Citations | Share | Domains |
|---|---|---|---|
| A practice's own website (architect, engineer, design-build, builder) | 174 | 63.3% | 49 |
| Directory or marketplace (Houzz, Angi, HomeAdvisor, Thumbtack and similar) | 28 | 10.2% | 9 |
| Professional association (AIA national and chapters, Boston Society for Architecture) | 25 | 9.1% | 4 |
| Media or trade editorial | 9 | 3.3% | 9 |
| Product manufacturer (siding, windows) | 9 | 3.3% | 2 |
| Academic paper or repository | 7 | 2.5% | 7 |
| Forum or social (Reddit, LinkedIn) | 6 | 2.2% | 3 |
| Unclassified | 17 | 6.2% | 14 |
Sit with the first two rows. The directory listings firms pay for and renew out of habit are one citation in ten. Peer firms writing about the hiring decision are six in ten. The marketplaces still matter for the reasons covered in Houzz for architects, but they are not what the answer is built out of.
The single most-cited domain in the whole set was one Los Angeles design-build practice, quoted 26 times across three of the four engines. It was not cited for its work. Every page of its that resolved to a real URL was a blog post, two of them on what an architect costs.
Every firm page cited was a guide, never a portfolio
This is the finding that should change what you publish. Here are real URLs the engines pulled from, exactly as they appeared in the citations:
/blog/architect-cost-for-renovations/post/how-to-choose-the-right-architect-for-your-home/blog/average-architect-fees-residential/choosing-commercial-architect-firm-for-a-new-build-out//how-to-choose-a-commercial-architecture-firm-for-your-project//things-to-consider-when-choosing-a-commercial-architect//education/how-to-choose-a-good-architect//how-to-hire-an-architect-for-a-custom-home/
Thirty-nine distinct firm URLs resolved across the four engines and all thirty-nine follow that shape. A language model answering a question about hiring an architect needs prose about hiring an architect. Your project pages are images and captions, which is right for a visitor and nothing at all for a machine. That gap is the same one behind the portfolio problem, arriving now in a second place.
None of these are big publishers. They are ordinary practices, several of them small, that happened to write the page. The bar is low right now and it will not stay low.
The two buyers get answered out of different libraries
Splitting the citations by question shows the fork clearly. The residential answers leaned on cost pages and choose-an-architect posts from residential firms, builders and design-build practices, plus two product manufacturers whose blogs happen to rank. Homeowners are being answered largely by people selling to homeowners.
The commercial answers came from a narrower and more institutional set. Every one of the 25 association citations sits on a commercial question, mostly AIA chapters, and the firm pages cited were explicitly commercial ones.
The commercial buyer is being handed fewer sources with more authority behind them, which is harder to break into and more valuable once you are there. The rest of that sale is in marketing for commercial architects, and the residential version in marketing for residential architects.
Google's AI Overview is already on these searches
Six of the seven buyer queries we pulled returned an AI Overview above the organic results: how to choose an architect, how much does an architect cost, how to hire an architect for a custom home, how to choose a commercial architect, questions to ask an architect, and architect fees for residential projects.
The one exception was "architect near me", which still runs on a three-slot local pack, which is why your Google Business Profile is still doing real work.
Two of those six Overviews returned their reference lists in full, giving 16 citations. Ten of the 16 were a practice's own website. Two were Reddit, two were AIA pages, one was Houzz and one was YouTube. And 11 of those 16 domains had already turned up in the separate chat-engine pull. Two different retrieval systems landing on nearly the same set of pages is the part worth trusting here.
The engine rewrites your client's question before it searches
Gemini reported the searches it actually ran. One question became three or four, and the rewrites are the phrasings your page should be built around: "questions to ask architect custom home", "architect fee structure custom home", "average architect fees percentage US", "factors to consider when choosing an architecture firm for commercial project", "key criteria for choosing an architect".
Those are section headings, not keywords. Answer each one plainly, near the top, in the words the engine used, and you have written a page that can be lifted whole. The method for turning buyer questions into pages is in content marketing for architects, and the technical layer that lets a crawler read any of it is in SEO for architects.
The honest limits of all this
Being cited is not the same as being visited. Pew Research Center analyzed the browsing of 900 US adults across 68,879 Google searches in March 2025 and found that when an AI summary appeared, people clicked a search result on 8% of visits against 15% without one, and clicked a link inside the summary on 1%. Getting named in the answer is the win. Expecting traffic from it is not.
There is a second limit worth stating. ChatGPT running gpt-4o returned zero cited sources on all four of our questions and answered from what it already knew, telling the residential buyer to ask friends, check the AIA and review portfolios. A meaningful share of AI answers about your firm never touch the live web at all, and nothing you publish this quarter will change those.
So treat this as the compounding layer rather than the engine. It works slowly and it does not fill next quarter. The demand you can actually schedule still comes from putting the work in front of the right people on purpose, which is Meta ads for architects, with this sitting on top of it.
One last thing worth knowing. Almost nobody is searching for this topic: "chatgpt for architects" runs about 70 US searches a month and "ai seo for architects" returns no volume data at all. Your clients are not searching for how AI works. They are asking it for an architect.
A few honest answers
How do I get my architecture firm recommended by ChatGPT?
Publish the pages the engines are already quoting: plain-prose guides on what your service costs, how to choose a firm like yours, and what to ask before hiring, each answering one question directly and near the top. Then make sure the page is static HTML a crawler can read without running JavaScript. There is no submission form and no way to pay for placement, which is the good news.
Do I need to be listed in directories for AI to find me?
Less than the industry assumes. Directories and marketplaces were 10% of the citations we logged, against 63% for practices' own sites. Keep the free listings current because they help a real person verify you exist, and stop treating a paid tier as the route into an AI answer.
Does this replace SEO for architects?
It is the same work aimed at a second reader. The pages that earn AI citations are the pages that rank, written more directly, and the technical requirements overlap almost entirely. What changes is the payoff to expect from each page, and how you measure whether it landed.
How would I even check whether an AI recommends my firm?
Ask it, in a logged-out session, the question a client would ask, and write down what it says and what it cites. Same day each month, same wording, so the answers are comparable. Vary the wording once a quarter to see how fragile the result is. It takes twenty minutes and it is the only measurement here that costs nothing.
Write the page your client is about to ask about
The uncomfortable read on our data is that a firm with a beautiful site and no writing on it is invisible to the systems its clients now use first, while a small practice that posted a fee explainer three years ago is quoted to those clients by four different engines. None of the cited firms did anything sophisticated. They answered the question in public.
Start with the two questions every client asks anyway: what this costs, and how to choose. Put real numbers in both, publish each as its own page, keep them current. The wider plan they belong to is in architecture marketing, and the pipeline they eventually feed is in lead generation for architects.
If you would rather have the whole system built and run for you while you stay on site visits, that is what we do, and we get paid when a qualified consultation actually shows up.
Book a call with IPQmedia