AI search visibility for contractors: get recommended when homeowners ask an AI

Open ChatGPT and ask it for the best roofer in your city. It will not hand you ten blue links. It will name three or four companies, in a confident paragraph, with no indication that it considered anyone else.

If your company is not one of those names, you were eliminated from the homeowner's shortlist before they ever visited a website. You will never see it in your analytics. There is no impression to count, no click that did not happen, no line in a report. The call simply goes somewhere else.

What actually changed

For twenty years the sequence was: search, scan a list, pick two or three, call. The list was the shortlist, and being on page one meant being in the running.

That sequence now often starts one step earlier. A homeowner asks ChatGPT, Perplexity, Gemini or Google's AI Overview a question in plain language, such as who should I call, what does this cost, or is this company any good. They get a synthesized answer with a handful of names in it. Some of them click through afterward. Many do not. Either way, the shortlist was assembled by a machine, and you were on it or you were not.

Being cited is not the same as ranking

This is the part that trips people up, including people who sell SEO. Ranking is a contest to appear in a list of links. Citation is a contest to be the source a model reaches for when it assembles a sentence.

The two overlap, since a page nobody can crawl gets neither, but they are scored differently. An engine building an answer pulls from what it can find and attribute: structured pages that state facts plainly, directory and review platforms, industry associations, and other people's content that happens to mention you. A page that ranks well because it is long and well-linked can still be useless to a model that needs one clean, attributable sentence about who serves Tampa.

What happened when we actually measured it

In August 2026, we ran the test rather than speculating about it. Five homeowner prompts per company, across ChatGPT, Perplexity, Gemini and Claude, for 13 roofing companies. That is 260 answers.

Twelve of thirteen never had their own website linked in any of their five answers. One company, O'LYN Roofing in Greater Boston, scored a single win out of five. That was the high score.

The names the engines handed out instead were not local competitors. They were manufacturer brands like GAF, which does not install anything, and national consolidators: Tecta America, CentiMark, DaBella, Erie Home, Power Home Remodeling.

The single most useful data point: one tested company ranked for 2,647 Google keywords and still appeared in zero AI answers. That is the whole argument in one line. Ranking and being cited are different problems, and solving one does not solve the other.

One more finding worth sitting with. Competitors' own "best roofers in [city]" blog posts were repeatedly cited by the engines as neutral authority: a company's marketing page, treated as an impartial recommendation. The full method, the per-company results and the raw prompts are in the full study.

What the work actually involves

There is no button for this, and anyone selling one is selling something else. The work is unglamorous and mostly consists of making your business legible to a machine that is trying to assemble a trustworthy sentence.

Answer-shaped content

Real homeowner questions, answered plainly, in a form that can be lifted out of the page and still make sense. "How much does a roof replacement cost in Tampa?" answered in three sentences with a real range beats 1,800 words that circle the question and end in a call to action.

Structure a machine can parse

FAQ blocks, service and organization schema, clean headings, and an llms.txt file that tells AI crawlers what your site contains and where the important pages live. This is plumbing, not magic, but skipping it means an engine has to guess at what your page says.

Service and city pages with genuine specifics

An engine answering "who does emergency AC repair in Fort Worth" needs a page that unambiguously establishes both the service and the city. Thin pages that swap a town name in and out do not just fail to rank. They give the model nothing to cite.

Consistent business facts everywhere

Name, address and phone identical across your site, Google Business Profile, and every directory that lists you. Conflicting facts across sources make a model less confident about naming you, and a model that is not confident picks someone else.

Reviews, in volume and recency

Review platforms are among the sources engines lean on hardest for local recommendations. Forty reviews from this year carry more weight than two hundred that stop in 2023.

Mentions in sources engines already trust

Directories, association listings, local press, supplier and manufacturer pages. Being named on third-party sources the engines already read is how you get into an answer that never touches your own website.

What this can and cannot do

This is weeks-to-months work, not an overnight switch. Content has to be published and indexed, third-party sources have to update, reviews accumulate at the speed of actual jobs.

You are also not going to outrank GAF. A shingle manufacturer with decades of brand authority will keep showing up, and that is fine. The homeowner still has to hire someone to install the thing. The realistic goal is being one of the handful of local names an engine returns alongside the giants.

Nobody can guarantee placement in an AI answer. The engines do not sell placement, do not publish their ranking logic, and change their behavior without notice. Anyone promising you a spot is lying.

What can be promised is the work, done properly, and a measurement you can check yourself with the same prompts, any time, for free.

How it is delivered and what it costs

AI-search visibility is not sold as a separate retainer here. It is part of how the site gets built in the first place, because retrofitting it onto a site that was structured badly costs more than doing it right once.

Scope and price are fixed in writing before work starts, and you own the domain, the site and the leads.

The honest version

Almost nobody selling contractor websites is doing this work yet, which is the only reason it is still cheap to be early. In a few years it will be table stakes and the companies that started in 2026 will be the ones already sitting in the answers.

If you want to know where you stand before spending anything, the free AI-visibility test runs the same five prompts on your company across four engines and sends you what came back. If you would rather read first, start with how to get recommended by ChatGPT. No call, no pitch, and the results are yours either way.