AI Visibility for Architecture Firms: Prompts, Wrong Facts and Fixes

Architecture buyers rarely type a firm name. They describe a building type, a style, a constraint or a place. Here are hypothetical examples of the prompts a developer, school board or homeowner might use:

• "Which architecture firms in Denver specialize in K-12 school design?"
• "Best residential architect near me for a modern farmhouse"
• "Architects in Portland with Passive House experience"
• "Who designed the new public library in [city]?"
• "Healthcare architects with outpatient clinic experience in the Southeast"
• "Is [firm name] licensed to practice in Texas?"
• "Compare [Firm A] and [Firm B] for a mixed-use project"
• "Mid-century modern renovation architect in Palm Springs"

Each prompt carries a sector, a style, a credential or a geography. An engine can only match your firm to those attributes if they appear as readable text somewhere it can fetch. Keep in mind that engines rewrite prompts. OpenAI says ChatGPT search rewrites a query into one or more targeted queries. So write each attribute as a plain fact, not one exact keyword phrase. Residential studios face a similar style-driven pattern, covered in AI visibility for interior design studios.

Different buyers use different engines, so check each one separately.

• Google AI Overviews and AI Mode catch "architect near me" and location searches. Google says a page must be indexed and eligible to show with a snippet to appear as a supporting link. There are no extra technical requirements.
• ChatGPT is where many clients research a shortlist. Its search answers depend on OAI-SearchBot being able to reach your site.
• Perplexity cites sources openly, which makes it useful for seeing which third-party pages shape your description.
• Microsoft Copilot grounds its answers in the Bing index. Being in Google tells you nothing about Bing.

As of October 2026, retrieval is discretionary on every platform. Sometimes an engine searches the web, and sometimes it answers from what the model already knows. That second path runs on frozen training data. OpenAI's model docs list a Feb 16, 2026 knowledge cutoff for the GPT-5.6 family and Aug 31, 2025 for the Instant path. A principal who joined in spring 2026 won't exist in those answers, whatever you publish. If one engine cites you and another ignores you, why Perplexity cites a site but ChatGPT doesn't explains the usual causes.

Architecture has three fact types that AI engines garble more than most.

Licensed principals

Principals move between firms, retire or found new studios. Old bios survive on award pages, conference sites and press releases. An engine may name a departed partner as your lead, or claim a license in a state where nobody at the firm is registered.

Sector focus

Firms shift sectors over a decade. A practice that moved from retail to healthcare can still be described as a retail firm, because older press coverage outnumbers newer pages.

Project attribution

This is the hardest one. A building can have a design architect, an architect of record, a joint venture partner and an interiors firm. Award listings and news stories often name only one. An engine may credit your project to a partner, or credit you with a partner's project.

Say you run a hypothetical 20-person firm that was architect of record on a hospital wing. The press release named only the design architect. An engine asked "who designed the hospital wing" will repeat the press release. The fix is a clear sentence on your project page stating your exact role, plus a request to the publisher to add your credit.

Architecture sites lead with photography, which is right for clients and weak for engines. An engine quotes sentences, not photos. A project page with twelve images and one line of caption gives it almost nothing to repeat.

Vercel's crawler study found ChatGPT and Claude crawlers fetched JavaScript files (11.50% and 23.84% of requests) but didn't execute them. Many portfolio themes load captions, project data and filters with JavaScript. Googlebot renders JavaScript. Most other AI crawlers don't.

For each project page, put these facts in the HTML before any script runs:

• Project name, city and completion year
• Building type and sector
• Your firm's exact role (design architect, architect of record, interiors)
• Size, budget band or certification, where you can publish it
• Partner firms and the principal in charge
• Awards won, with the awarding body and year

A specific, fetchable fact gives the engine a sentence it can quote. That is the only positive experimental signal for on-page content in the evidence. To test, view the page source with JavaScript off. If the project description is missing, AI crawlers likely see it missing too.

When engines cite other sites instead of yours, being named correctly on those sites is the route into the answer. The evidence shows a strong association here, not a guaranteed effect. For architects, check these pages for facts that match your own site:

• Award pages: AIA chapter awards, design competitions and sector awards. Check the firm name, your role and the principal credited.
• Directories: Google Business Profile, Houzz, AIA member listings and local chamber directories. Check address, services and sectors.
• State licensing board lookups: confirm which principals hold active licenses and where.
• Design press: ArchDaily, Dezeen and regional business journals. Look for missing or wrong role credits.
• Client and partner pages: the school district, hospital or developer page for the project, plus engineer and contractor case studies.

Make a simple sheet: fact, your version, each third-party version. Correct your site first, then ask each publisher to update theirs. To see which rival firms engines name for your sectors, try the free competitor AI visibility check.

Access is where the evidence is strongest. These checks remove blocks. They don't add rankings.

1. robots.txt per bot. Each AI crawler has its own rule. OpenAI states its settings are independent: OAI-SearchBot controls ChatGPT search, and GPTBot governs training. A blanket AI block often removes ChatGPT search eligibility by accident. See whether to block GPTBot for the trade-off.
2. Firewall and CDN. OpenAI says eligibility requires allowing OAI-SearchBot and confirming your host or CDN allows its published IP addresses. A firewall can return 403 to a crawler that robots.txt allows. The AI Crawler Checker requests a page as each AI crawler and shows which ones get turned away.
3. Snippet controls. Search your templates for noindex, nosnippet, max-snippet:0 and data-nosnippet. Google and Bing document these as controls that keep content out of AI answers. Portfolio themes sometimes apply them to gallery pages.
4. Index status in two places. Check Google Search Console and Bing Webmaster Tools separately.

A minimal robots.txt that allows search crawlers while opting out of OpenAI training looks like this:

User-agent: OAI-SearchBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Claude-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

Days 1 to 7: baseline

Write eight prompts that match your sectors, styles and cities. Run them in ChatGPT, Google, Perplexity and Copilot. Record whether your firm appears and which facts are wrong. The free checker on this site gives a snapshot for three buyer questions. The guide to checking brand visibility in AI search covers the manual method.

Days 8 to 14: access

Run the robots.txt, firewall, snippet and index checks above. Fix any block before touching content.

Days 15 to 21: project pages

Add plain-text descriptions to your top fifteen project pages. State role, sector, location, year and principal in charge. Update principal bios with current licenses.

Days 22 to 30: third parties and re-check

Send corrections to award, directory and press pages. Re-run the same prompts and compare. Track results over months with the method in measuring AI share of voice. Expect slow, uneven change, because retrieval and model updates are outside your control.