Home › Guides

How roofers get recommended by ChatGPT, Gemini, Grok, Claude and Perplexity

Short answer: In our test of 40 US cities, ChatGPT and Perplexity gave the same top roofer pick only 11% of the time. In 56% of those city-by-city comparisons, their lists had no roofer in common. The $49 report therefore asks all 5 leading AIs: ChatGPT, Gemini, Grok, Claude and Perplexity.

A storm rolls through, and a homeowner with a leak doesn't scroll through ten search results. They ask an AI assistant who to call, and they call one of the two or three names it gives them. Here's what those AI tools look at when someone asks for a roofer, what we saw when we tested it, and why good roofing companies often get left out.

From our 960-answer study (October 25, 2026): 240 roofer answers, using ChatGPT and Perplexity, in 40 US cities.

How an assistant chooses a roofer

ChatGPT, Gemini, Grok, Claude and Perplexity cannot inspect a roof or judge a crew's workmanship. A homeowner who asks "who's the best roofer near me?" or "who fixes hail damage in my town?" gets names drawn from what those assistants can read online, usually the companies with the clearest, most consistent information.

Skill on the job is not enough on its own. The roofer who appears in the answer is the one the assistant can find and make sense of.

The questions homeowners actually type

Roofing searches tend to be narrow and time-sensitive. Typical wording includes:

• "Who can fix a roof leak today in [town]?"
• "Best roofer for storm or hail damage near me"
• "Roofing company that works with insurance claims"
• "Metal roof installers in [county]"
• "Family-owned roofer, not a franchise"

Every wording is a separate chance to appear, or to be skipped. A company can be named for "roof replacement" and still be absent for "emergency roof repair" or "insurance claim help."

What 240 roofer answers actually showed

As part of the 960-answer study, ChatGPT and Perplexity were asked "Who is the best roofer in [city]?" in 40 US cities, three times each, producing 240 roofer answers. These points stood out:

• The list stays short. ChatGPT named 4.5 roofers per answer on average, and Perplexity named 2.2. In 24% of its roofer answers, Perplexity named just one.
• The two AIs often disagreed. For the same city, ChatGPT and Perplexity gave the same top pick 11% of the time, and in 56% of cases their lists had no roofer in common.
• Asked three times, the top name often moved. ChatGPT kept the same top pick all three times in just 1 of the 40 cities, and across those three answers it named 10.6 different roofers on average. Perplexity kept the same top pick all three times in 5 of 40.
• BBB came up the most. Perplexity cited BBB in 92% of its roofer answers (110 of 120). ChatGPT cited it in 29%.
• Roofing "best of" lists and Angi mattered. ChatGPT's most-cited site for roofers was a roofing ranking site, theroofingbrief.com, in 34% of its answers (Perplexity: 28%). Angi showed up in 20% of ChatGPT's roofer answers and 23% of Perplexity's.
• Manufacturer contractor lists showed up. Perplexity cited GAF's website in 22% of its roofer answers, and manufacturer or trade-association directories in 23%.
• Company websites mattered for Perplexity. Perplexity pulled up roofing company websites in 90% of its answers. ChatGPT linked to a company's own website in 9%.
• Licensing sites, too. Perplexity cited a state or city licensing or government site, such as California's contractor license board, in 25% of its roofer answers.

Citations below are the websites the assistants opened, not the roofing companies they named, and a citation is not an endorsement. We grouped site types from the web addresses in each answer and counted a company website when the address belonged to the business.

The roofer numbers come from two assistants, one question, and one day. Gemini, Grok and Claude may use other sources, so the paid report covers all five.

The information assistants tend to use

In the roofer questions we run, the same kinds of sources come up again and again:

• Google Business Profile. Category, services, the towns you cover, hours and photos.
• Reviews that describe real jobs. Recent notes about a hail replacement or a leak repair give an assistant more to work with than a stack of old "great job!" lines.
• Directories and manufacturer lists. BBB, Angi, Yelp and contractor locators from manufacturers showed up often in our study.
• The company website. Clear wording about the work you do, where you work, licensing, insurance, warranties and whether you offer inspections.
• Matching contact details. The same business name, phone and address on every profile.

Why a solid roofing company still gets missed

These are patterns from the roofer study, not a verdict on your company. Whether any of them apply depends on your site, your listings, and the other roofers nearby.

• The website is closed to automated visitors. A "checking your browser" screen can stop assistants even though homeowners never see it.
• Every service sits on one vague page. Separate sections for repairs, replacements, storm damage and the roof types you install give an assistant more to match than "all roofing services."
• Towns served are left unsaid. If the cities and suburbs you cover are not written down, an assistant may not tie you to a homeowner there.
• Reviews are thin or stale. Recent reviews that describe the actual job tell the assistant more than generic praise from years ago.
• What makes you different is never spelled out. Family-owned, years in business, manufacturer-certified, help with insurance claims: if it is not in plain words, the assistant cannot repeat it.

A review request you can send after the next job

After each finished job, send the homeowner a short text with your Google review link and ask them to mention what you did and the town, for example "replaced our roof after the hail storm in Plano." Reviews like that give AI the kind of detail it uses. Never buy reviews.

Why the first fix is not the same for every roofer

Two roofing shops on the same street can be missing for unrelated reasons. One site may be unreachable to assistants. Another may appear for replacements and never for storm damage. A third may lose every question to a franchise with far more directory listings. Starting in the wrong place costs time, and a guess will not show whether the answers moved.

The AI Recommendation Report ($49) is how you see your own case. We run 15 real homeowner questions about roofers in your area through ChatGPT, Gemini, Grok, Claude and Perplexity, twice each, and show how often you are named next to up to 3 local competitors, which websites each assistant used, and the Top 5 Fixes for your company. The Report + Fix Kit ($99) adds ready-to-paste wording and step-by-step instructions.

To see the layout, open our sample report, built from an insurance-agency test: who was named, which sites were used, and what to change first. We cannot promise an assistant will name you, because answers change, but this is how you learn where you stand.

Questions about the roofer test or the report? Email Christine at hello@beaioptimized.com.

Other trade guides from the same study: insurance agents · electricians · pest control · concrete contractors · plumbers · dentists · HVAC companies · the full 960-answer study.

Questions roofers ask

Results for other trades: concrete contractors · dentists · electricians · flooring companies · HVAC companies · insurance agents · painters · pest control · plumbers · tree service · drain cleaners · window replacement · the full 960-answer study.