• AI visibility

Guide · AI search

Will architects and contractors find our products when they ask AI what to specify?

They can, if the performance, compliance and sustainability facts behind your products are published where assistants and specifiers can verify them. Architects still use AI sparingly for product research, but they call it one of their least efficient tasks, and AI referrals to product research platforms are growing fast.

By Underneath Team Guide Updated

This guide is for manufacturers of building materials and building products: envelope, roofing, insulation, structural, concrete, interiors, finishes and fixtures. These companies win business when an architect or engineer writes a product into a specification, when a contractor buys it at bid time, and when a distributor keeps it on the shelf. The goal of AI visibility here is more specifications and better inquiries, not website traffic.

The short version

  1. AI use among architects is early: only 6% regularly used AI tools in their practice, according to the American Institute of Architects’ Journey to Specification research, as reported by Architect magazine (opens in a new tab) and Dezeen (opens in a new tab), although 53.1% had experimented with it.
  2. Product research is where the opportunity sits: architects in that study named product research and updating product lists among their most inefficient tasks, yet fewer than 10% of firms had tried AI on those tasks.
  3. Specifiers look for proof on origin and sustainability: in the 2026 edition of the AIA study (opens in a new tab), 72% of architects preferred US-manufactured products, and the share that proactively recommend sustainability options rose from 58% in 2020 to 79% in 2025.
  4. AI referrals are rising on product research platforms: ArchiPro (opens in a new tab), which serves New Zealand and Australia, recorded 69,192 sessions from AI platforms in the second quarter of 2026, up 662% on a year earlier.
  5. Every specification feeds a very large build market: US Census Bureau figures show construction spending running at an annual rate of $2,203.1 billion in August 2026.

Who decides which building materials get used, and what is one specification worth?

Architects and engineers write the specification, contractors buy against it, and distributors supply it.

A building product usually has to win more than once. First, a specifier, such as an architect, specification writer, interior designer or engineer, names it as the basis of design. Then a contractor or installer prices it at bid time and may propose an “or equal” alternative. Finally, a distributor or dealer supplies it, and on residential projects the homeowner often has a say. Each of those people researches products, and each can now ask an assistant.

The work they choose materials for is substantial. In August 2026, the Census Bureau estimated annual rates of $882.3 billion for residential construction, $773.0 billion for private nonresidential construction and $547.8 billion for public construction.

Price pressure is real at the contractor stage. In the AGC and Sage 2026 Construction Hiring and Business Outlook, 53% of contractors named materials costs among their biggest concerns for 2026. Makers of drones, robots and sensors meet the same buyers; see how jobsite technology reaches contractor shortlists.

No public figure shows what one specification is worth, because order values depend on the product and project. Our inference: the value is in what follows. A product written into a specification is priced by every bidding contractor, and one that earns a place in a firm’s master specification can recur across many later projects. A product that is never considered is never priced at all.

How much are specifiers, contractors and homeowners using AI today?

Lightly among architects, more among contractors and homeowners, and growing fastest in early product discovery.

Among architects, the Journey to Specification study, run by AIA with Deltek and ConstructConnect and answered by more than 500 AIA-registered architects, found only 8% of firms had built AI into their processes. Most architects who did use AI were using chatbots such as ChatGPT, grammar tools and image generators. A majority were optimistic about AI for complex problems (84%), but nearly all (90%) had concerns, led by inaccuracy.

Contractors have moved faster. In the 2026 Houzz State of AI in Construction and Design Report (opens in a new tab), 52% of construction firms used AI for everyday business tasks. The same habit shapes how contractors pick construction software. Homeowners are experimenting too: 22% of renovating homeowners had used AI tools for their projects, and among them 50% used it to compare options.

Engineers who specify structural, mechanical and electrical products follow a similar pattern. In the 2026 State of Marketing to Engineers (opens in a new tab) from TREW Marketing and GlobalSpec, 69% of technical buyers used generative AI during the purchasing process. The same survey found 76% routinely researched in online technical publications, slightly more than the 74% who used supplier and vendor websites.

The clearest product-discovery signal comes from ArchiPro, a platform where architects, builders and homeowners research building products in New Zealand and Australia. Its AI referrals generated 174,729 pageviews in the quarter, up 745% on a year earlier. ArchiPro notes that AI traffic is still relatively small. It is not US data, but it shows the direction of travel.

What do specifiers and builders ask AI about materials?

Questions that combine performance, code compliance, sustainability, origin and the project type.

We wrote the following examples to illustrate the pattern. They were not collected from real users, and we have not tested what any assistant answers to them.

NeedIllustrative question
Basis of design“Continuous insulation options for a steel-stud wall in a four-story building that meets the fire test for exterior walls”
Code evidence“Fiber cement cladding with a current ICC-ES evaluation report”
Sustainability“Lower-carbon concrete mixes with product-specific environmental declarations available near Denver”
Origin“Roof membranes made in the US for a Buy America project”
Substitution“What is an approved equal to the specified acoustic ceiling panel for a school?”
Residential“Composite or PVC decking for a coastal home with salt spray”

Every one of these questions turns on documents a manufacturer controls: test reports, evaluation reports, listings, environmental declarations, origin statements, warranties and installation instructions. If those exist only inside gated downloads or scanned PDFs, an assistant may not be able to read them, and a specifier in a hurry may move on.

How does an AI answer turn into a specification, a bid and an order?

Through early research, a basis-of-design choice, bidding, a submittal, then an order through distribution.

  1. Researched. During design, an architect or engineer looks for products that meet performance, code and sustainability goals, using manufacturer sites, research platforms, representatives, peers and, increasingly, AI.
  2. Specified. The product is named in the project specification, sometimes with listed alternatives.
  3. Bid. Contractors price it. Some propose substitutions, which the design team accepts or rejects.
  4. Submitted and approved. The contractor submits product data for approval before ordering.
  5. Ordered and installed. The order goes through a distributor or dealer.
  6. Repeated. A product that performs may be kept in the office’s master specification for future projects. That is our inference about how value compounds, not a measured rate.

The inquiry also tends to land on the manufacturer’s own site. In ArchiPro’s quarter, visitors who clicked through to suppliers’ websites generated 11,014 inquiries there, against 2,512 sent through the platform itself. Our inference: AI visibility mostly shows its value in steps one and two, and the evidence of it may arrive as an inquiry or a specification that no report attributes to AI.

What makes an assistant name one product and not another?

Verifiable, consistent facts in sources it trusts; the platforms describe their search process, not their choices.

Documented by the platforms. OpenAI says (opens in a new tab) ChatGPT search typically rewrites a question into one or more targeted queries sent to search partners, and that sites must allow OAI-SearchBot to be eligible. Google says (opens in a new tab) AI Overviews and AI Mode may use a “query fan-out” across subtopics, and that a page must be indexed and eligible to show a snippet to be a supporting link. Neither explains how products are chosen.

Observed in our studies. In our study of assistants’ hidden searches, ChatGPT searched for a named publication, ranking or award in 43.8% of its answers. In our study of business facts in AI answers, phone numbers differed from the business’s profile 30.6% of the time when that number was missing from the business’s own website, against 1.6% when it was there. The study measured local businesses, not products, but the lesson carries: when your own sources disagree, answers drift.

Observed in specifier research. The 2026 AIA study found 73% of architects highly value being involved in product development by manufacturers, while only 24% want to take part directly. Specifiers want manufacturers who understand their problems, and evidence of that, such as published project collaborations, is something an assistant can find.

Our inference for building materials. The trust factors are the documents specifiers already check, published so they can be read and matched: performance data with the test standard and lab named, current code evaluation reports and listings, environmental declarations and origin statements, installation and warranty terms, completed projects naming the architect, and trade-press or award coverage. ArchiPro makes a similar point, arguing that products connected to completed projects, professionals and technical documentation give AI richer context than isolated product pages.

What does a manufacturer lose when AI answers leave its products out?

It loses consideration at the design stage, where the specification, and often the order, is decided.

We found no measurement of specifications lost to AI, so we label our reasoning:

  • Early exclusion is hard to reverse. If a product is not considered at design, the contractor prices whatever was specified, and a substitution request has to argue against it. That is our inference about how specification works.
  • Preferences are shifting toward proof. With 72% of architects preferring US-made products and 79% proactively recommending sustainable options, missing origin or environmental facts can remove a product from a shortlist even when it qualifies.
  • Wrong facts are worse than none. An assistant that repeats an outdated fire rating, a withdrawn evaluation report or a discontinued color can rule a product out. Our guide to fixing wrong brand information in AI answers explains where to start.
  • The loss leaves no trace. A specifier who never sees your product sends no inquiry. Our guide to why analytics miss AI visibility explains why such losses rarely show up in reports.

How does GEO work for a building materials company?

Generative engine optimization (GEO) makes your products easy for assistants to find, describe accurately and support with independent evidence.

For a building materials manufacturer, the work usually includes:

  1. Readable product data. Performance values, test standards, sizes, colors and limitations on product pages in plain text, not only inside data-sheet PDFs.
  2. Open technical files. Specification text, CAD details and BIM objects available without a sign-up wall.
  3. Current approvals. Evaluation reports, listings and certifications with their numbers, scope and dates, matching the issuing body’s own record.
  4. Sustainability and origin facts. Environmental declarations, recycled content and where each product is made, stated per product rather than per company.
  5. Consistent channel data. The same names, specifications and approvals on distributor, dealer and research-platform listings as on your site, so assistants do not meet conflicting facts.
  6. Built proof. Project pages that name the building, the architect and the products used, published with permission.
  7. Independent coverage. Trade publications, continuing-education courses, awards and association programs, the named sources assistants search for. Our guide on how brands build authority for AI search covers the approach.
  8. Measurement. Ask a fixed set of performance, code, sustainability and substitution questions across ChatGPT, Gemini, Perplexity, Copilot and Google’s AI features, repeat them, and track which products and sources appear. Compare the trend with inquiries and specification wins.

No manufacturer can make an assistant write its product into an answer. What this work can do is make your product the easiest one for a specifier, or an assistant, to check. Manufacturers that also sell made-to-order components can compare notes with our guide for contract manufacturers, and our guide on product content AI shopping assistants prefer is useful for retail and residential lines.

What can’t the specification data tell us yet?

It shows research habits and AI curiosity, not how many specifications AI answers have changed.

  • No attribution. None of the sources links an AI answer to a specification, bid or order.
  • The AI finding is dated. The 6% figure comes from an AIA study reported in March 2025; adoption may have moved since.
  • Interested parties. ConstructConnect and Deltek sell to manufacturers and architects; ArchiPro sells listings to suppliers; Houzz runs a platform for building professionals; GlobalSpec sells advertising.
  • Different markets. ArchiPro’s data covers New Zealand and Australia, not the US.
  • Our studies did not test material questions. Applying their findings to specification is our inference.

Where should a building materials company start?

Start by asking assistants the questions a specifier asks before naming a product, and see what they say about yours.

That first check usually shows whether your products appear for their performance, code and sustainability questions, whether the facts given match your current data, which platforms, publications and distributor listings the answers rely on, and which competing products appear instead. The work then is to publish your product facts in readable, consistent form and earn the independent coverage that confirms them.

If more of your revenue should start with a basis-of-design listing, arrange a review of your products’ AI visibility with us. We test the performance, code and substitution questions specifiers ask in your category, trace which data sheets, listings and publications the answers draw on, and set out the product facts to publish so more specifications and qualified inquiries follow. Our generative engine optimization service page explains how that work runs for a manufacturer, from readable data sheets and current approvals to matching distributor listings.

Frequently asked questions

Do architects use ChatGPT to find building products?

Some do, but regular use was low: 6% of architects in the AIA study used AI regularly, mostly chatbots, grammar tools and image generators. Product research is one of the tasks they find least efficient, which is where AI use could grow.

Are BIM objects and specification text still worth producing?

Yes. Specifiers use them in design, and published, ungated files are also evidence an assistant can find. Keep them current and consistent with your product pages.

Should we put our data sheets behind a form?

Gating can collect leads, but it can also hide your facts from assistants and from specifiers in a hurry. One middle path is to keep core technical data open and gate only deeper tools.

Does AI visibility matter if we sell through distributors?

Yes, because the choice of product is often made before the distributor is involved. Keep distributor listings consistent with your own data so answers do not contradict you.

Sources

Free strategy call

Some questions are easier to answer about your own business.

Bring the one that matters most. On a free 30-minute call we’ll take a first look at it and send you a short written read afterward.

Ask about my businessHow we work