Does ChatGPT recommend specific contractors?

Yes, often. In WorkBrag's small September 2026 test across ChatGPT, Gemini and Claude, 55% of answers named at least one specific contractor. The rest gave a checklist of what to look for instead of a name.

What do AI engines look at before naming a contractor?

Mostly public proof: Google Business Profile reviews, Yelp and other review sites, Reddit threads, YouTube videos and the contractor's own site. In our test, 8 of 10 named businesses had 50 or more Google reviews and one from the last 60 days. None of the three engines cited a paid ad.

Do ChatGPT, Gemini and Claude cite the same sources?

Rarely. Orbit Media's 2026 study of 13,184 citations found the engines agreed on the same domain only 1.7% of the time. ChatGPT leans on primary sources, Gemini on Reddit and YouTube, Claude on review sites and never Reddit.

← Research

AI search

We asked ChatGPT, Gemini and Claude for a Dallas roofer

By Asaf Shamly, founder · Last updated September 24, 2026 · 6 min read

55% of the time, an AI engine handed us the name of a specific Dallas-Fort Worth contractor. In WorkBrag's own small test of 20 prompts across ChatGPT, Gemini and Claude on September 18-19, 2026, 8 of 10 named businesses had 50 or more Google reviews and a review from the last 60 days, and none of the three engines cited a paid ad.

Homeowners now ask chatbots for a roofer the way they used to ask a neighbor. BrightLocal's 2026 Local Consumer Review Survey puts ChatGPT and other AI tools at 45% of consumers as a place they check reviews, up from 6% a year earlier, now third behind Google and Facebook. So we ran a test. Not a study, a test.

Say you run a two-truck roofing outfit in Mesquite. You've got 31 Google reviews, the last one from April, and a Facebook page you haven't touched since the Super Bowl. The Tuesday after a hailstorm, a homeowner three streets from your last job asks Gemini who to call. Your name doesn't come up, and you'll never know it didn't.

How did we run the test?

On September 18 and 19, 2026, WorkBrag typed 20 homeowner-style prompts into ChatGPT, Gemini and Claude. Prompts like "who is a good roofer in Plano for a hail claim" and "best reviewed roofing company in Fort Worth, not a storm chaser." Same wording in all three engines, 60 answers total.

1.7%

of the time, the AI engines agreed on the same domain

Source: Orbit Media's 2026 study of 13,184 citations

Logged for each answer: whether a business was named, what sources the engine cited and what the named businesses looked like when we checked their public profiles by hand. We're not naming those businesses here. The point is the pattern, not the roster.

What surprised us while we sat there typing: how often an answer read like a Google profile's review section with the names filed off, and how much more careful all three engines got on the storm-chaser prompt. We'll rerun the same 20 prompts, same wording, the third week of every month, and change nothing so the numbers compare.

Small sample, one metro, run once. We will rerun it monthly and update this page. Treat the numbers below as a snapshot, not a benchmark.

Do AI engines name a specific contractor?

In 55% of answers, yes. The other 45% gave a vetting checklist: check the license, get three quotes, read recent reviews. Gemini named a business most often. Claude was the most cautious.

And we'll say this flatly: the checklist answers are the better outcome for a good roofer. A homeowner told to look for a license, insurance and recent reviews walks straight into your strongest material. The Mesquite roofer with nothing since April is the one who should dread that list.

When an engine did name a business, the profile behind it was rarely thin. 8 of 10 named businesses had 50 or more Google reviews and at least one review from the last 60 days. 6 of 10 had an active Facebook or Instagram page the engine cited. None of the 60 answers cited a paid ad of any kind. Not one.

Each engine leaned on a different kind of proof to name a roofer

Source: WorkBrag internal test, 20 prompts x 3 engines, Dallas-Fort Worth, September 18-19, 2026. Shares are rounded. Small sample, one metro, run once; not a modeled estimate and not a benchmark.

Why do ChatGPT, Gemini and Claude cite different things?

They are built by different companies with different reading habits, and the bigger studies back up what we saw at the keyboard. Orbit Media's 2026 study of 13,184 citations found the engines agreed on the same domain only 1.7% of the time.

Gemini favors Reddit and listicles. ChatGPT favors primary sources and vendor pages. Claude favors review sites and never cited Reddit.

Count matters too. Orbit Media puts it at about 4.5 sources per ChatGPT answer, 8.1 for Gemini and 3.6 for Claude. A Gemini answer is a wide net. A Claude answer is a short list of pages it trusts, and if you're not on one of them you don't exist.

Our 60 answers matched that. ChatGPT leaned on Google Business Profile and Yelp pages. Gemini leaned on Reddit threads and YouTube. Claude leaned on review sites and the contractor's own website. Peec AI's running tally adds that Reddit sits at number one or two among cited domains on all five major AI engines.

What did the named contractors have in common?

Volume, recency and a live social page. Nobody got named for the biggest truck fleet or the slickest website. The engines went for businesses other people had recently written about in public places a machine could read.

Stale is the killer. Omnibound's 2026 roundup reports that 76% of ChatGPT's most-cited pages were updated in the last 30 days. A review stream that stops in March looks abandoned to a chatbot in September, even if the crew worked every week in between.

What this means for a roofer

You can't buy your way into a chatbot answer. You earn your way in, and the work looks like this:

Get past 50 Google reviews and keep at least one landing every month. Volume gets you considered; recency keeps you there.
Keep Facebook or Instagram alive with real job posts. Six of ten named businesses in our test had one.
Put a short customer video on YouTube and your own site. Gemini reads YouTube, Claude reads your site.
Ask every customer, not just the ones you remember. This is the job WorkBrag does for you: one text after the job, the review and the video come back, and the clip is posted for you. Then check this page next month and see if the pattern held.

Methodology

Third-party figures come from BrightLocal's Local Consumer Review Survey 2026, Orbit Media's 2026 analysis of 13,184 AI citations, Peec AI's AI search statistics page and Omnibound's 2026 generative engine optimization roundup.

The engine test is WorkBrag's own: 20 prompts, 3 engines, 60 answers, Dallas-Fort Worth, run once on September 18-19, 2026, with named businesses checked by hand against their public profiles.

It is a small sample and we report it as such. Figures marked as WorkBrag modeled estimates are projections built from public benchmarks and our own launch assumptions, not measured results. We will replace them with measured data from live accounts and update this page.

Frequently asked questions

Does ChatGPT recommend specific contractors?: Yes, often. In WorkBrag's small September 2026 test across ChatGPT, Gemini and Claude, 55% of answers named at least one specific contractor. The rest gave a checklist of what to look for instead of a name.
What do AI engines look at before naming a contractor?: Mostly public proof: Google Business Profile reviews, Yelp and other review sites, Reddit threads, YouTube videos and the contractor's own site. In our test, 8 of 10 named businesses had 50 or more Google reviews and one from the last 60 days. None of the three engines cited a paid ad.
Do ChatGPT, Gemini and Claude cite the same sources?: Rarely. Orbit Media's 2026 study of 13,184 citations found the engines agreed on the same domain only 1.7% of the time. ChatGPT leans on primary sources, Gemini on Reddit and YouTube, Claude on review sites and never Reddit.