Will ChatGPT recommend your roofing company? Run this 10-minute check.

Most roofing companies are not named when a homeowner asks ChatGPT who to call. That is the normal result right now, and it does not mean something is broken on your website. You can confirm it yourself in about ten minutes with four questions and three free tools. Below are the exact questions to type, what each possible answer tells you, and an honest separation between what is actually known about getting named by these systems and what the industry is currently guessing.

What you need before you start

A phone or a laptop, ten minutes, and three places to type: ChatGPT, a plain Google search, and Perplexity. Free accounts are enough for all three. Run the searches in a private or incognito window if you can. Your own browsing history bends what comes back, and the point of this exercise is to see what a stranger in your service area sees.

Have a notepad open. Four questions across three tools is twelve answers, and you will not remember them.

The test

1. Ask for the best roofer in your city. Type: best roofing company in [your city]. Read the entire answer, not the first sentence. Write down every company name that appears.
2. Ask the emergency question. Type: who should I call for a roof leak in [your city]. This is closer to how a homeowner actually types at 11pm with a bucket on the floor. The answer often looks different from question one.
3. Ask the money question. Type: average cost of a roof replacement in [your city]. Nobody is hiring anyone at this stage, which is exactly why it matters. This is the question that gets asked first, and whoever the model names here gets a head start.
4. Ask about yourself, cold. Type your company name alone, with nothing else attached. Then follow up with: what do you know about [your company name] in [your city]. You are checking two things: whether the model has anything on you at all, and whether what it has is correct.
5. Run all four in Google. Look for the AI-generated block above the normal results. Note whether it appears, who it mentions, and which websites it links out to on the right side or beneath the text.
6. Run all four in Perplexity. Perplexity numbers its sources inline, which makes it the easiest of the three to audit. Click two or three of those numbers and see where the answer actually came from.
7. Log three things for every answer. Were you named. Which competitors were named. What sources were cited, and could you realistically get onto one of them.

How to read your result

Nobody local is named and you get generic advice

Common, and it tells you almost nothing about your business. It means the model did not find sources it trusted enough to name specific companies for that city. Your competitors are in the same position you are.

Competitors are named and you are not

This is the result worth looking at closely, but not for the reason you would expect. Go find where those names came from. In most cases it is a directory page, a “best roofers in [city]” listicle from a local publication, a Reddit thread, or a review aggregator. The company did not get named because of something clever on its own website. It got named because it appears on pages the model already reads.

You are named

Good, and hard to attribute. There is currently no reliable way to trace a mention back to a specific thing you did, and anyone who tells you otherwise is selling something. Screenshot it and re-run the same prompt in ninety days. These answers move.

The cold company-name question returns nothing, or returns something wrong

Nothing is normal for a private local company. Something wrong is the only genuinely urgent finding in this whole test. If the model has your service area, your phone number, or your specialty incorrect, that is a data problem sitting in public sources, and it is fixable through the same channels that feed Google Business Profile and the major directories.

What actually moves this, and what the industry is guessing about

Three findings here are measured. Everything after them is opinion, including opinions held confidently by people charging money for them.

Known. Whitespark ran 540 local queries manually and found AI Overviews appearing in 68% of local searches, against 39% for the traditional local pack. For informational local queries, 92% triggered one. For hybrid-intent queries, the kind that mix research and buying, 97% did; their own example was “average cost of dental implants in Phoenix,” which is structurally identical to question three above. The write-up is on Search Engine Journal.

Known. Kevin Indig analyzed 1.2 million ChatGPT responses and found that 44.2% of citations come from the first 30% of a page’s content. The practical version: if your roof replacement cost page answers the question in paragraph nine, after the company history and the mission statement, the part that answers it may never be read. Put the answer at the top.

Known. Omniscient Digital looked at more than 23,000 citations and found that on branded queries, only about 23% of citations point to the brand’s own website. Roughly 77% come from somewhere else. Reddit alone accounts for about 21% of AI Overview citations and YouTube about 18.8%. Your website is one input among many, and it is not the largest one.

Guessed. That specific schema markup gets you cited. That an llms.txt file on your server does anything. That publishing more often helps. That there is a format, word count, or heading structure that these systems prefer. Some of this may turn out to be true. None of it has been demonstrated with published methodology, and the confident advice you see about it is pattern-matching from regular SEO applied to systems that work differently.

Does this matter yet for a roofing company

Here is the tension, and it does not resolve cleanly. In the United States, “how to show up on chatgpt” gets about 10 searches a month. Every variation of it, including “how to get my business to show up on chatgpt,” measures at zero. Almost nobody is looking for this.

At the same time, 97% of hybrid-intent local queries now produce an AI-generated answer above the normal results. The surface is already in front of homeowners even though nearly nobody is searching for how to appear on it.

Both numbers are real. The honest reading is that the answer surface arrived before anyone figured out what to do about it, and no one has published data connecting an AI citation to a signed roofing contract. So do not rebuild your marketing around this. Also do not assume it stays at zero. The reasonable position is to run the test twice a year, fix anything factually wrong about your business in public sources, and keep investing in the things that help you either way: real reviews, pages that answer the actual question in the first paragraph, and presence on the local pages and threads that these systems already cite.

If you want a second set of eyes on your result

Run the test first. It costs ten minutes and you own the answer either way. If what comes back concerns you, or you want someone to look at where your competitors are getting named from and whether any of it is reachable, book a call here. Bring your notes from the table above.

Related: roofing SEO work.