ChatGPT Citations for Roofers
When someone asks ChatGPT for a roofer in their city, it answers with a handful of named companies. Being one of them has nothing to do with paying for placement and only partly to do with your Google ranking. It comes down to whether the sources ChatGPT reads can tell it, clearly, that you do this work in that place.
Get my free roofing SEO auditRoofing is the kind of purchase people ask about before they act. It is expensive, urgent when it goes wrong, and most homeowners have never hired a roofer before. Historically they asked a neighbour. Increasingly they ask an assistant.
The questions look like the ones they used to put into Google, but phrased as questions to a person: who should I call for hail damage in my area, is this company reputable, what should a replacement cost here, who handles insurance claims.
In September 2026, BrightLocal analyzed around 200,000 localized AI searches across roughly 1,300 businesses. In that dataset, the tracked business was mentioned in 33% of ChatGPT responses, compared with 66% of Google Maps searches. BrightLocal analyzed.
Two things follow from that. ChatGPT is a smaller discovery surface than Maps, so it should not replace local SEO. But it is a surface where very few roofing contractors have done anything deliberate, which makes it cheap ground to take.
ChatGPT answers local questions from two places, and the difference matters.
Live web results. For anything location-specific or current, it searches the web and cites what it used. These citations are visible, and they are where a roofing contractor can realistically appear.
Training data. Everything it absorbed before its cutoff. No citations, no recency, and nothing you can influence in the short term.
For “best roofing company in Austin,” it will almost always search. What it reaches for is a mix of directories and review aggregators, local news and community discussion, and business websites themselves. In the BrightLocal dataset, business websites accounted for 93% of unique source domains and 42% of all citations.
The important difference from Google is that ChatGPT is not ordering a list. It is assembling an answer from several sources at once, and it tends to name businesses that more than one source agrees on. A contractor who appears consistently across their own site, their Google Business Profile and a few directories is easier to assemble an answer around than one who exists in only one place.
The crawlers can't reach the site. OpenAI uses separate crawlers for different jobs — GPTBot for training data, OAI-SearchBot for search results, and ChatGPT-User for pages fetched during a conversation. Blocking them in robots.txt removes a site from consideration entirely. This is worth checking rather than assuming, because some site builders and SEO plugins add these blocks by default and the owner never sees it.
Nothing on the page states the obvious. A roofing site that says “quality workmanship since 1998” and “serving the greater metro area” has told a reader very little and a machine nothing. Which cities. Which services. Which roof types. Written as plain sentences, not implied by a page title.
There's nothing corroborating it. A website on its own is one source. A business that also has a consistent Google Business Profile, real reviews and a few legitimate directory listings has several sources saying the same thing, which is what an assembled answer is built from.
No evidence of the actual work. Stock photography of someone else's roof tells a reader nothing and gives an assistant nothing to cite. Project descriptions with the roof type, the problem and the outcome are specific enough to quote.
Confirm the crawlers can reach you. We check robots.txt and server rules for blocks on OpenAI's crawlers and remove them where they aren't intentional.
Rewrite the pages so the answers are extractable. Which cities, which services, which roof systems, stated plainly and early. Headings that match how a homeowner actually asks. FAQ markup on genuine questions.
Build the corroboration. A Google Business Profile that matches the site exactly, consistent name, address and phone across directories, and a steady flow of real reviews. This is the same work that moves the map pack, which is why it is worth doing regardless.
Publish evidence of the work. Real projects with the roof type, the problem and the result, with customer permission. Specific enough to be quoted, concrete enough for a homeowner to judge.
Add the structured data. Organization and Service markup connecting the site to the verified business, so the entity behind the content is unambiguous.
We can't guarantee a citation. ChatGPT's answers vary between runs and OpenAI changes how it sources them. What we can do is make you the kind of business an answer can be assembled around, and track whether it starts happening.