How to Get Your Roofing Company Recommended by ChatGPT

A homeowner with a leak used to open Google and start clicking. Some of them now type the question into ChatGPT and take the three names it gives back. Whether you are one of those three names is not luck, and it is not about being the best roofer in town.

I tested this on thirteen roofing companies in August 2026. Twelve of them never had their own website linked in any answer across five homeowner prompts and four AI engines. The full data is in the 13-company test. This post is the other half: what to actually do about it.

Step one: run the test on yourself

Before you spend a dollar, spend thirty minutes. Open ChatGPT, Perplexity, Gemini and Claude, and run the same five questions through each. That is twenty answers.

Score each answer on one thing: was your own website linked in the answer? Not whether your name showed up somewhere in the paragraph, but whether the answer pointed at your site. Those are different outcomes and only one of them sends you a phone call.

Then log two more columns: which companies won, and which websites the answer was built from. That second column is the important one. It is a list of exactly which pages on the internet decide who gets recommended in your market, and it becomes your to-do list.

Why AI answers work the way they do

An AI assistant answering a local question is not ranking contractors. It is running a search, reading a handful of pages, and assembling a summary it can defend with sources. Everything follows from that.

It means the engine has to find something specific enough to quote and attribute. A homepage that says quality workmanship since 1998 and serving the greater metro area gives it nothing. A page that says we replace tile roofs in Clearwater, typical replacement runs, say, 18,000 to 34,000 dollars, we pull the permit and here is what the county inspection covers gives it a sentence it can lift and a source it can name.

It also means most of the material comes from somewhere other than your website. In the study, answers leaned heavily on BBB, Angi, Expertise.com, HomeGuide, Thumbtack, HomeAdvisor, Birdeye and Trustpilot, and repeatedly on blog posts published by competing roofers. In Wichita, the best-roofing-companies list the engines kept citing was written by a competitor.

The work, in the order I would do it

One page per service, one page per city

If you do roof replacement, storm damage repair, gutters and metal roofing across six towns, that is a real page for each service and a real page for each town. Not six copies of the same paragraph with the city name swapped, which every engine and every homeowner can spot instantly.

Real means specific. What a roof costs in that town and why. Which permits that county requires. How fast you get there after a hailstorm. A job you did on a street somebody local would recognize, with photos. The one link any company earned in my study was to a specific service page, a gutters page, not a homepage. Two other companies had their pages read by an engine and still were not linked, because the page gave it nothing to hand to the homeowner.

Write your FAQs as actual questions

AI answers are built out of question-and-answer shaped text, because that is what maps cleanly onto what the user asked. So write in that shape. Take the ten questions your office manager answers on the phone every week and put them on the site as questions, each with a two-to-four-sentence answer directly underneath.

Answer them so they stand alone. An engine will lift one paragraph out of your page and show it with no context around it. If the answer only makes sense after reading the three paragraphs above it, it will not get used.

Reviews: volume, recency, and everywhere

In the Phoenix market I tested, a competitor picked up five separate mentions largely on the back of a 4.9-star average across 1,056 reviews. That number was doing the persuading, and the engines found it through review platforms, not through Google alone.

So ask every customer, every job, on the day the job is finished, and spread them across Google, BBB and Angi rather than pouring everything into one profile. Ask for honest reviews only: the FTC's 2024 rule bans bought and fake ones, and Google's review policy removes them. Old reviews decay in usefulness. A steady trickle beats a burst of forty from two years ago.

Get mentioned in other people's content

This is the part most contractors skip and it is the part the data pointed at hardest: every source an engine cites is somebody else's page. You want to be on those pages.

• Manufacturer contractor directories for every brand you are certified on.
• Local chamber, trade association and supplier partner listings.
• Local news and TV segments. After a storm, reporters need a roofer to quote, and that is a phone call you can make.
• Podcasts, local business interviews, home-show recaps, anything with a published page.
• Your own city guides. If competitors are writing the best roofers in [city] posts that AI treats as neutral authority, there is no reason you cannot write an honest one too.

llms.txt and FAQ schema, without the jargon

[FAQ schema](https://schema.org/FAQPage) is a small block of hidden code that labels your questions as questions and your answers as answers, so a machine reading the page does not have to guess. Most roofing sites do not have it. It takes a developer about an hour.

[llms.txt](https://llmstxt.org/) is a plain text file at the root of your site, at yoursite.com/llms.txt, that lists your key pages and describes what each one covers, in the same spirit as a sitemap but written for AI crawlers. It is cheap to add. No engine has said publicly that it reads the file, so treat it as insurance, not a lever.

Be honest about what these do. They are plumbing, not authority. In my study one company had FAQ schema and still scored zero. It only had three ranking keywords, so there was nothing for the schema to help. Fix the content first, then make it machine-readable.

Keep your business facts identical everywhere

Name, address, phone, hours, service area, license number. Same on your site, Google Business Profile, BBB, Angi, Facebook, every directory. When an engine finds three different phone numbers for you, it has no clean fact to state and reaches for a competitor it can.

This is boring and it matters. Spend an afternoon with a spreadsheet and fix every listing you can find, including the dead ones from a previous marketing company.

How long this takes

Weeks to months. Nobody should tell you otherwise.

• Weeks 1 to 2: run the self-test, fix listings and business facts, claim review profiles, add llms.txt and FAQ schema, start asking for reviews on every job.
• Weeks 3 to 8: build out the service and city pages with real specifics, and write the FAQ content. This is the bulk of the work.
• Months 2 to 6: earn outside mentions, keep reviews flowing, publish local guides. Re-run the twenty-answer test monthly and watch which columns change.

The lag is not the engines being slow. New pages have to be crawled and indexed, third-party mentions accumulate at their own pace, and different assistants refresh their sources on different schedules. You are changing what exists on the internet about your company, and that takes as long as it takes.

What this will not do

It will not put you above GAF or a national consolidator on every prompt. Those brands have content volume you are not going to out-publish, and for nine of the thirteen companies I tested they were the default answer.

What it will do is get you into the local slots, the two or three actual contractors an engine names alongside the national brands, and get your own page cited when it does. That is the realistic ceiling, and right now almost nobody in your market is competing for it.

If you want a starting point, the free 31-point website audit covers most of the on-site items above, and you can work through it yourself in an afternoon. If you would rather see where you currently stand, send me your company name and I will run the same five-prompt test across all four engines and send you the raw answers, free. If you want that test re-run monthly with the work done for you, that is the AI search visibility service. And if the honest answer is that the site itself is the bottleneck, I build fixed-price: Starter Site from $1,450, full builds from $2,925, live in one to two weeks.