August 17, 2026

/ AEO

11 min read

GEO for architecture firms in 2026: winning AI project discovery queries

AI engines cannot read a portfolio of renderings. Here is how architecture firms turn visual project work into citable text and win AI queries in 2026.

GEO for architecture firms in 2026: winning AI project discovery queries

GEO for architects in 2026 means converting a visual portfolio into retrievable text, because ChatGPT, Perplexity, and Google AI Overviews cite sentences, not renderings. The firms winning project discovery queries publish project pages that state typology, square footage, budget band, timeline, and jurisdiction in plain language, then reinforce them with licensure signals from NCARB and the AIA plus press coverage in ArchDaily, Dezeen, and Architectural Record. The opening is wide: Archmark reviewed more than 5,500 architecture firm websites and found only about 25 percent were set up to be indexed properly, while BrightEdge measured AI Overviews on roughly 48 percent of tracked Google queries as of February 2026.

Three more numbers explain why this matters right now. NCARB By the Numbers 2025 counts roughly 116,000 licensed architects in the United States, a 4 percent drop in a single year, so the pool of firms competing for each project is shrinking while the discovery layer changes underneath them. The AIA Architecture Billings Index has sat below 50 every month since January 2026, printing 47.3 in June and 44.5 in May, which means billings are contracting and new work is scarcer than it was two years ago. Nonresidential building construction spending fell 2 percent to $846 billion in 2025 and dropped another 7 percent through the first five months of 2026. Softer demand plus a smaller licensed population equals a market where the firm that gets named inside a Gemini, Microsoft Copilot, or Google AI Mode answer takes the meeting, and the rest never learn the project existed.

The problem is structural. Architecture sells through images, and images are the format AI retrieval handles worst. Houzz, Architizer, Angi, Instagram, and Pinterest have built enormous text layers around architectural work, so when someone asks an assistant a question about hiring an architect, those platforms get quoted and the practices that did the work do not.

Which architecture queries do AI assistants actually answer?

Four clusters carry hiring intent, and every one of them has a numeric or factual answer that a practicing architect already knows. None of them are being answered by firms today, which is exactly why the citations go to directories.

1. Cost and fee structure queries

“How much does an architect cost,” “what percentage do architects charge,” “architect fees for a home addition.” Design fees typically run 8 to 15 percent of construction cost, with a wider band of 5 to 20 percent depending on complexity, and hourly rates land between $100 and $300 for licensed staff while principals bill $200 to $400 in major markets. Angi, HomeGuide, and Houzz publish those ranges constantly. Architects almost never do, and the citation follows the publisher.

2. Scope and credential queries

“Do I need an architect or a designer,” “when does a permit require a stamped drawing,” “architect vs draftsman.” These are jurisdiction-dependent questions with real answers. A designer cannot stamp permit documents; a licensed architect or structural engineer can. Some states exempt single-family homes and residential buildings up to four units from the architect requirement entirely, and a firm that publishes the exemption rules for its own state and county owns a query no national directory can answer as well.

3. Local discovery queries

“Best architects in Charleston,” “modern residential architects near me,” “firms that do adaptive reuse in Denver.” Assistants build these lists from Google Business Profile, Houzz, Yelp, AIA chapter directories, and local press. If the firm is missing from any of those, it is missing from the list.

4. Typology and process queries

“How long does ADU permitting take,” “what happens in design development,” “how much does a dental office buildout cost per square foot.” Process questions map cleanly onto the AIA phase structure of schematic design, design development, and construction documents, and every phase has a duration a firm can state honestly.

Most architecture firms have never checked whether an assistant names them when a homeowner asks who designs modern residential work in their city. Get your free AI visibility audit and see which project discovery prompts your firm already appears in.

Why does an award winning portfolio produce zero AI citations?

Because retrieval systems index text and a portfolio is a wall of images. Forty photographs, a project name like “Harbor House,” and a one-line caption contain no extractable claim: no typology, no square footage, no budget, no constraint, no decision.

The second failure is sameness. Every architecture site has a projects grid, and to a crawler they look structurally identical. Nothing distinguishes the firm that does historic tax credit rehabs from the firm that does ground-up multifamily from the firm that does high-end coastal residential, so no engine can match a specific query to a specific practice.

The third failure is distribution. Many firms publish their strongest work to Instagram, Architizer, and Houzz before it ever appears on their own domain, which means the indexable version of the project lives somewhere the firm does not control. Interior designers hit the same wall, and the fix is the one covered in our GEO guide for interior designers.

How do you turn a project into a page an engine can quote?

Write each project as a labeled text record, not a gallery with a caption. Seven fields do the work, and a principal can dictate all seven in under ten minutes per project.

Typology and use. Name it the way a client would search: single-family new construction, ADU, adaptive reuse of a warehouse, ambulatory surgery center, K through 12 addition, ground-up multifamily.

Square footage and program. State the number and the unit or room count. “18,400 square feet, 22 units, 4,100 square feet of ground floor retail” is quotable. “A thoughtful mixed-use intervention” is not.

Budget band. Publish a range, not an invoice. “Construction cost in the $4M to $6M range” gives an engine something to anchor on and qualifies a prospect before the first call.

Timeline in months by phase. Schematic design through permit through substantial completion. Permitting duration in your jurisdiction is one of the most searched and least published numbers in the industry.

Jurisdiction and code path. City, county, zoning district, and any variance, historic review, or coastal setback approval you navigated. This is what makes a firm the obvious answer to “who has done a historic review project in this district.”

Named consultants. Structural engineer, MEP, civil, landscape, general contractor, and lighting designer, by firm name. Named entities are what retrieval systems match against, and every consultant you name is a relationship an engine can corroborate.

Outcome. Occupancy achieved, energy performance, LEED or Passive House certification, appraised value, award received, or press earned.

Seven short paragraphs of that beat seventy photographs for citation purposes. The photographs still matter for conversion once a human lands on the page. They just do not do the retrieval work.

How does commercial project discovery differ from residential?

Completely, and firms that serve both should build two separate content tracks. Residential clients are individuals researching alone, and they ask cost, timeline, and scope questions before they ever type a firm name. Commercial clients are developers, facility directors, and public agencies running a structured process.

On the residential side the winning assets are a fee page, a permitting timeline page for your county, an ADU or addition guide, a Google Business Profile with the Architect category and 30 or more current photos, and an active Houzz profile.

On the commercial side the buying process is qualifications-based. Public agencies operate under the Brooks Act and most states mirror it, meaning firms are ranked on demonstrated competence through an RFQ before fees are discussed. What an owner or a procurement lead now does before writing the shortlist is ask ChatGPT or Microsoft Copilot which firms in the region have done this building type. Your answer to that is a sector page: healthcare, industrial, civic, hospitality, education, each carrying three named past projects with square footage, delivery method, and completion year. Add LinkedIn activity from named principals and coverage in the local business journal, and you become retrievable at the moment the list is formed.

Which entity signals tell AI engines your firm is real?

Licensure is the strongest trust signal architecture has, and almost nobody publishes it in machine-readable form. Put the firm’s state registration number, each principal’s license jurisdictions, the NCARB Certificate status, and AIA membership with the specific chapter on a single credentials page, in text, not in a logo strip.

Then make the firm consistent everywhere it exists: Google Business Profile, Houzz, Architizer, Archinect, Yelp, LinkedIn, and the AIA chapter firm directory should all carry an identical name, address, phone number, and category. Awards are entity fuel too. AIA Honor Awards, AIA chapter design awards, Architizer A+ Awards, and Architectural Record Design Vanguard recognition are verifiable third-party facts an engine can cross-check, and engines trust what multiple independent sources agree on.

Which press and directories actually move AI citations?

Architectural press punches far above its traffic numbers because AI engines weight editorial sources heavily. A single ArchDaily project feature typically produces a high-authority backlink plus a permanent, well-structured text description of a project that engines can quote for years. Dezeen, Architectural Record, Metropolis, Architizer, and Archinect work the same way.

Local coverage matters just as much for local queries. A city business journal story about a project breaking ground, a preservation society writeup, or a chamber feature all become corroborating sources an assistant leans on when someone asks who designs in a specific market. Pair those with directory presence on Houzz, BuildZoom, Porch, and Angi for residential intent, and Google Business Profile for everything. Third-party accounts of your work always outrank your own claims about it.

What schema markup should an architecture firm run?

Five types cover the entire site. Mark the firm as ProfessionalService nested under LocalBusiness, with areaServed, priceRange, and address populated. Mark each principal as Person with hasCredential for the license and memberOf for the AIA. Mark each project page as CreativeWork with dateCreated, contentLocation, and a plain-text description carrying the seven fields above. Mark every rendering and photograph as ImageObject with a caption, contentLocation, and creator credit so the visual work carries retrievable metadata. Mark cost and process pages as FAQPage.

None of it works if the underlying text is missing. Schema describes content, it does not create it. Implementation details are in our schema markup guide for AI search.

Your renderings are doing nothing for you inside an AI answer. Get your free AI visibility audit and find out which of your project types and markets are already invisible to ChatGPT and Perplexity.

FAQ

How long does GEO take to work for an architecture firm? Expect 60 to 90 days for the first citations on low-competition queries such as permitting timelines and county-specific scope questions, and 4 to 8 months for competitive local prompts like “best architects in Austin.” Firms with existing ArchDaily or Architectural Record coverage move faster because the corroborating sources already exist. Firms starting with no press and a thin Google Business Profile should plan on the longer end.

Do architects need a Google Business Profile if most work comes from referrals? Yes. Google Business Profile is a primary input to local AI answers across Google AI Overviews, Google AI Mode, and Gemini, and it is also what a referred prospect checks before calling. Use the Architect category, load 30 or more current project photos with descriptive filenames, keep hours and service area accurate, and answer the Q and A section yourself with real permitting and fee questions.

Should we publish our fees when the industry norm is not to? Publish ranges, not fee schedules. Stating that design fees typically run 8 to 15 percent of construction cost on your project types, with a stated minimum project size, wins the highest-volume query in the category and filters unqualified inquiries before they reach a principal. Angi, HomeGuide, and Houzz already publish those numbers. The only question is whether the citation points at them or at you.

Does posting to Instagram and Pinterest help AI visibility? Indirectly and weakly. Both platforms drive human discovery and neither gets cited often in AI answers, because the text layer is thin and the content sits behind platform structure. Publish the project on your own domain first with full text, then push images to Instagram, Pinterest, Houzz, and Architizer with a link back. Treat social as distribution, not as the canonical record of your work.

How do commercial architecture firms measure GEO results? Track named-mention rate across ChatGPT, Perplexity, Gemini, and Microsoft Copilot for your sector and market prompts, RFQ invitations received, and referral traffic from AI sources in GA4. Commercial cycles are long, so the leading indicator is shortlist appearance rather than form fills. A firm that goes from zero to a named mention in three of five sector prompts has moved, even before the pipeline reflects it.

Is GEO different from the SEO our web developer already does? Yes. SEO optimizes for a ranked list of links and rewards keyword coverage and backlinks. GEO optimizes for being quoted inside a generated answer and rewards extractable facts, named entities, structured data, and third-party corroboration. A page can rank fifth in Google and be the source an AI Overview quotes, or rank first and never be cited. We break the distinction down in GEO vs SEO.

The firms that win will not be the best designers

They will be the ones whose work is legible to a machine. That is uncomfortable to say about a discipline built on craft, but the mechanics are not negotiable: an assistant assembling a shortlist can only name a firm it can find, verify, and quote. The average practice publishes a projects grid, a two-paragraph about page, and a contact form, which gives an engine nothing while Houzz and Angi answer the questions that firm’s own clients are asking.

Billings have been below 50 all year and nonresidential spending is still falling. In that market, every project awarded without your firm on the shortlist is a project you never had a chance to lose. Pick your five strongest projects, write the seven fields for each, publish your fee range and your county’s permitting timeline, and get your license and AIA credentials into text on a real page. That is a week of work, and it puts you in a conversation your competitors have not noticed is happening.

Tagged

geo architecture ai search aeo local seo