Architect AI Search Citations — the Project-Page Playbook

The work is concentrated where the AI-citation gravity is strongest. Project pages do most of the heavy lifting because architecture queries are typology-and-place-shaped — AI engines preferentially cite specific project pages that name both. The Person entity does the second-largest share because it disambiguates the Registered Architect from a designer, an interior decorator, or a contractor — a distinction that matters legally in all 55 US jurisdictions⁷ and matters to AI engines that weight verifiable credentials.

The realistic 2026 ambition for a boutique firm is to be cited on the typology + place queries within the firm’s typology specialism and region. “Adaptive reuse warehouse Brooklyn” produces a Perplexity answer that cites two to four sources; if the firm’s project page on a completed warehouse adaptation in Brooklyn opens with a 134-167 word lead naming the typology, the location, the building’s prior use, and the design strategy, the page is competitive for one of those citation slots. The traffic that follows is small but high-intent: someone researching the typology becomes a candidate client by the time they reach the firm’s contact form.

The generic queries are largely unreachable in AI search and the firm should not optimise for them. ChatGPT and AI Overviews explicitly decline to recommend a specific architect on most “best architect for X” queries — they route the user to ArchDaily, to the AIA “Find an Architect” directory, to the regional design press, or to the local pack. Optimising the Squarespace site for “best residential architect” through the lens of AI citation is a misallocation of effort; that ranking lives in the directory and publication layer, which a separate publication-feature strategy handles.

The pattern is testable. Search Perplexity for “adaptive reuse warehouse Brooklyn” and inspect the citations. The top citation is typically an ArchDaily project page where the typology and the borough appear in the first paragraph. The second and third citations split between firm sites whose project page leads with the same vocabulary, regional design publications (Brownstoner, Curbed New York, Architect’s Newspaper), and AIA New York chapter award announcements. The pages that lose are the ones whose openers are about the firm rather than about the project — engines cannot extract a typology + place signal from a paragraph that does not contain one.

For a boutique Squarespace site, the implication is direct. Every project page on the firm’s site should open with a self-contained passage that names the typology (passive house, adaptive reuse, mid-century renovation, mass timber commercial, net-zero single family), the location (the city, the neighbourhood, the region), the building (its prior use, its size, its programme), and the lead architect. The passage stands alone — an AI engine can lift it without context and the lift still reads as a complete answer to the query. That is the citation surface.

The CreativeWork⁵ block does specific work on architecture project pages. It carries name (the project name), description (a 60-80 word real description, not boilerplate), image (the hero shot as an ImageObject), creator (a reference back to the Person Registered Architect via @id), locationCreated (the city), and dateCreated (the year). The graph pattern means AI engines walk from the project page to the firm to the architect to the credential set in a single reasoning step — which is exactly the path that produces a confident citation on typology + place queries.

 <script type="application/ld+json"> { "@context": "https://schema.org", "@type": "CreativeWork", "name": "Maple Street House", "description": "Passive-house renovation of a 1924 craftsman bungalow in Burlington, Vermont. R-40 double-stud envelope, triple-glazed fenestration, HERS-15 rating.", "image": "https://yourstudio.com/assets/maple-street-house-hero.jpg", "dateCreated": "2024", "locationCreated": { "@type": "Place", "name": "Burlington, VT" }, "creator": {"@id": "https://yourstudio.com/#principal"}, "about": [ "Passive House Design", "Cold-Climate Residential Architecture", "Adaptive Reuse" ] } </script> 

The creator reference resolves to the Person Registered Architect declared on the homepage’s LocalBusiness + Person graph (the production block in the architects hub). The about array carries the typology vocabulary that AI engines match against the typology + place query. The block is small, but the join is what carries the weight — engines walk from the project page through the creator reference to the architect to the firm to the credential set, and the walk is what produces the citation.

The four layers join the firm, the architect, and the project to verifiable third-party sources. Schema.org’s Person specification⁴ describes knowsAbout as “indicates topics a person has knowledge about, suggesting possible expertise but not implying it” — the wording matters because knowsAbout is a self-asserted signal, but AI engines weight it more heavily when the same expertise is verifiable through the sameAs URLs (AIA directory listing of the typologies the firm is known for, ArchDaily feature on a project in the typology, AIA chapter award announcement). The combination of self-assertion plus third-party verification is what produces the disambiguation; either alone is weaker.

For multi-state firms, the hasCredential array lists each jurisdiction-specific Registered Architect designation separately (“Registered Architect — Vermont”, “Registered Architect — New York”, “Registered Architect — Massachusetts”) plus the NCARB Certificate where held. The memberOf array lists the national AIA plus each state AIA chapter the architect is active in. The sameAs array links to the AIA member directory entry, the state board licensee lookup for each jurisdiction (most state boards offer a public “verify a license” URL), and any ArchDaily or Dezeen feature URLs. The graph compounds over years — every additional feature, every additional jurisdiction, every additional verified credential lifts the trust signal on every page that joins through @id.

The eight-to-twelve query benchmark is the spine. Pick the typology + place combinations that match the firm’s actual practice — not aspirational queries the firm hopes to grow into, but queries on completed projects the firm can cite. Examples for a Vermont passive-house specialist: “passive house architect Vermont”, “cold-climate residential architect Burlington”, “net-zero renovation Vermont”, “mass timber residential Vermont”, “passive house renovation 1920s bungalow”, “adaptive reuse barn Vermont”. The first three are firm-positioning queries; the last three are project-specific queries that should cite the firm’s most photographed project pages by month six.

The six-week check cadence aligns with this site’s general verification cycle for AI-search pages — the AI engine policies, citation patterns, and surface behaviour move fast enough that monthly is too slow on the leading edge and quarterly misses material drift¹. The log captures which engine cited (ChatGPT, Perplexity, AIO, Gemini, Copilot), which query, which firm-site URL was the citation source, and what paragraph the engine quoted (often visible in Perplexity, sometimes visible in ChatGPT, rarely visible in AIO). The paragraph capture is what feeds the next-cycle editorial change — if the engine quotes a paragraph the firm thinks is weak, that is the paragraph that gets rewritten first.