By the time a general contractor hears about a project, the list has usually already been cut. That was true before any of this. What changed is who does the cutting, and how early.
What is actually measured, and what is not
Start with the limitation, because it is the honest place to start: nobody has published a construction-specific version of this research. There is no survey of 500 owners and developers telling you what share of them asked an AI assistant to name general contractors for a 90,000 square foot medical office fit-out in Tarrant County. If someone quotes you that number, they made it up.
What exists is the general business-buying population, measured twice, by serious people.
Forrester's Buyers' Journey Survey — roughly 18,000 global business buyers, published January 2026 — found that 94% of business buyers used AI somewhere in their most recent buying process. The same research puts the average buying group at 13 internal stakeholders plus 9 external participants, and found that buyers are more likely to engage a provider based on information from industry experts than from AI tools.
Semrush surveyed 643 U.S. B2B professionals in March and April 2026, keeping 622 valid responses, of whom 519 were confirmed AI users. Fifty-four percent were final decision-makers. The stage-by-stage findings:
| Stage of the buying process | Share using AI |
|---|---|
| Early research and category scoping | 72% |
| Actively comparing vendors | 62% |
| Narrowing the shortlist | 48% |
| Supporting the final decision | 45% |
And the outcome questions:
- 97% said AI had helped them discover a vendor they did not previously know. For 44%, frequently.
- 92% said AI had shaped their vendor shortlist. For 45%, significantly.
- 83% said AI influenced their final decision. For 32%, majorly.
- 89% expect to rely on AI more for work decisions going forward. Fewer than 1% expect to rely on it less.
The sample skews toward technology, manufacturing, and professional services; construction is not broken out. So treat the exact percentages as a read on business buyers generally, not on your next owner. The mechanism is what transfers, and the mechanism is not industry-specific: a person with a budget, a committee, and a deadline asks a machine to narrow the field before they call anyone.
The number that should change how a mid-sized GC thinks
Asked what made a vendor stand out in an AI recommendation, buyers said:
Seven percent. In the surface that is now building the shortlist, being a name people recognize is nearly worthless, and being the precise answer to the question asked is worth more than seven times as much.
That inverts the economics every mid-sized general contractor has lived under for thirty years. The reason the $200M firm wins the interview it should not have won is reputation, relationship, and the fact that the owner's architect has heard of them. None of those three things exists inside a language model's first pass. What exists there is whether your published material — your site, your profile, the registries and directories that describe you — contains an unambiguous, machine-legible match for this delivery method, this project type, this market, this constraint.
A $20M contractor who has written a genuinely specific page about tenant finish in occupied medical space in Tarrant County can beat a $400M contractor whose website says "we deliver excellence across a diverse portfolio of market sectors." Not because the small firm is better. Because the big firm's page is not an answer to anything.
The AI does not award the contract
It is worth being precise about what the research says, because the overclaim here is easy and wrong.
Buyers do not read an AI answer and sign. They read it and then go verify. From the same survey, after an AI recommendation:
- 71% visit the vendor's website
- 63% search Google for the company
- 46% compare against alternatives
- 38% check third-party review platforms
Forrester's read is consistent: buyers validate AI output against peers, product experts, and analysts, and more than 60% buy some form of trial before committing. Their framing is that buyers are risk-averse and want proof, not promises.
So the sequence is: AI builds the list. Verification narrows it. People close it. Your relationships, your safety record, your references, your interview team — all of that still decides the job. It just decides it among the three names that made the list.
Which means the question for a commercial contractor is not how do I get an AI to sell for me. It is the much simpler and more uncomfortable one: am I on the list at all, and if not, why not?
Why a qualified contractor gets skipped
Almost never because a model judged the work. Three mechanical reasons, in the order they have to be fixed:
Eligibility. The crawler could not reach the page, the content rendered client-side and arrived as an empty shell, or the robots file blocked the agent outright. This is binary and it is the first thing to check, because nothing downstream matters if the answer here is no. Whether your site is blocking AI crawlers takes about ten minutes to determine.
Entity confidence. The engines do not rank you, they resolve you — assembling one picture of who you are from your website, your Google Business Profile, license and registration data, and the directories that carry you. When those sources disagree about your legal name, your trade classification, or your service area, confidence falls and the system names a company it is more certain about. That is the local entity gap, and in commercial work it is worse than in residential, because GC entity data is scattered across state registries, prequalification platforms, and plan rooms that nobody in the company has looked at in four years.
Specificity. This is the one the 53% figure is about. The answer layer rewards material that resolves a particular question, which is the opposite of how capability pages are usually written. Building an answer-ready page and the question-cluster model are the structural fixes; the short version is that one substantial page about a specific delivery method in a specific market will out-earn eight pages of "markets we serve."
There is also a buyer-side confirmation of all of this in the data: 25% of buyers said their chief frustration with AI recommendations was that relevant vendors were missing. The gap is not only visible from your side. The owners already notice it.
Find out whether you make the list.
The AI Visibility Audit runs your company through all six layers — GEO, AEO, LEO, VEO, PEO, IEO — and tells you whether the engines name you when a buyer asks who to call in your market, plus the two things worth fixing first. $27, self-serve, no call.
Get My AI Visibility Audit — $27The hour-long version of this test
You can establish where you stand today without buying anything.
- Write five sentences the way a buyer would actually type them. Not keywords. Sentences. "Who are the best general contractors in Fort Worth for a ground-up self-storage build?" "Which Dallas GCs do occupied medical office renovations?" "I need a CM-at-risk firm for a municipal project in Tarrant County." Include delivery method, project type, market, and constraint — because that is how the question arrives.
- Run all five through ChatGPT, Gemini, Perplexity, and Google's AI surface. Same wording, same day.
- Record three things per run: were you named, were you cited with a link, and which companies were named instead.
- Look at the sources the engines used. This is the step people skip and it is the most informative one. If the answer is built from a directory, a registry, or a plan room rather than from company websites, that tells you where the work is. We mapped the source ecosystem behind AI recommendations and found the same third-party surfaces doing the heavy lifting again and again — the same pattern we documented in the supplier signal.
- Run it again in two weeks, with the same five sentences. One check is an anecdote. The same check repeated is a measurement. How to test whether AI recommends your business has the full protocol.
Expect the results to be uneven across engines, because they always are. In our own tracked data across 700 checks in one Texas market, the same company hit a 92% mention rate on one engine and zero on another inside the same window, on identical work.
Why this is a window and not a crisis
Two clocks again. The engines improve on a six-month cycle. Commercial contractors improve their published material on a multi-year one, if at all — and a real share of them are still shipping sites that cannot be parsed and profiles that contradict their own license records.
Meanwhile 68% of Google searches now end without a click, which means the surface where the shortlist gets built is getting more decisive, not less, and 89% of the buyers in that survey expect to lean on it harder next year than this one.
There are ten blue links on a results page and roughly three names in an answer. That is a harder room to get into and a much better room to be in. Right now it is also a mostly empty one, because the firms with the budget to dominate it are the same firms whose 7% brand-recognition advantage does not carry across the threshold.
If you want the whole framework, it is on the method page and in the AEO field manual. If you want the commercial-specific version, that is what we do for commercial contractors. And if the thing actually keeping you up is not the next pursuit but what the last twelve are costing you, the Contractor Profit Leak Audit is the other half of the same problem — a number that exists, arriving somewhere it can be acted on.
Sources
- Forrester Research. Buyers' Journey Survey, fielded 2025, published January 2026; nearly 18,000 global business buyers. 94% reported using AI during their buying process; average buying group of 13 internal stakeholders and 9 external participants; more than 60% purchased some form of trial. The State of Business Buying.
- Semrush. "How AI tools shape the B2B buying process: A survey of 600+ US business professionals," 2026. 643 surveyed March–April 2026, 622 valid responses, 519 confirmed AI users; 54% final decision-makers. Source of the 97% / 92% / 83% discovery, shortlist and decision figures, the stage-by-stage usage, the standout factors including the 7% brand-recognition figure, the post-recommendation verification behavior, and the 25% missing-vendors frustration. Sample skews technology (20%), manufacturing (15%) and professional services (14%); construction is not reported separately. Survey.
- SparkToro. "In 2026, Less Than One Third of Google Searches Still Send a Click," analysis of Similarweb U.S. clickstream, January–April 2026. Study. Our contractor reading is at 68% of Google Searches Now End Without a Click.
- The 700-check figures are our own tracked data from one Texas market, published with methodology at We Ran 700 AI Visibility Checks in One Texas Market.
- No construction-specific survey of owner or developer AI use exists as of this writing. Every application of the B2B findings to commercial contracting is ours and is reasoning by mechanism, not measurement.