Be the roofer they find. Roofing SEO and AI search for roofers

A homeowner with a leak asks a search engine or an AI assistant, and each gives its own list of roofers. We asked Claude and ChatGPT for a roofer 945 times, in 63 UK towns and cities, and counted who they named. Here is what the answers drew on, and the work we do for a roofing firm.

What we found, and what we do about it.

The study asked each assistant one everyday question about each town, five times. Between them Claude and ChatGPT named 1,229 firms, counted town by town, and 1,090 of those were named by only one of the two. Four findings matter most to a roofer.

  1. Put the same firm on every list.

    Most of the pages both assistants cited were trade directories, quote sites and review sites. trustatrader.com was cited in 72% of Claude’s answers and checkatrade.com in 69% of ChatGPT’s. A firm cannot write its own reviews. It can make sure every listing says the same true thing.

  2. Hold Google’s map as well.

    Google shows a map with three firms above its results. ChatGPT never named 82% of the firms on that map, and Claude never named 48%. A place on one list does not carry over to the other, so the work covers both.

  3. Say where you are, so a machine can read it.

    The study’s question named a town and no country. Asked through the developer interface, 8% of the answers named firms in a town of the same name abroad. Nobody can stop a machine taking one Birmingham for another. A roofer’s site can say its town, its county and its country plainly, in its words and in its structured data, so a machine that reads it knows which one.

  4. Count it every month.

    Ask again and the list changes. Most of the firms ChatGPT named were missing from at least one of its five answers, so one answer on one day proves little. We ask every month, five times each, about your towns and your kinds of work, and count how often the firm is named.

A roofing company, in the numbers.

Mammoth Roofing is a roofing company with crews across Texas. We built its website and did its search work.

Nothing on the site claims what the company cannot show. A storm notice appears on a town’s page only when the National Weather Service reports a storm near it, each of the 95 town pages carries that town’s own storm history, and town pages without local evidence of the company’s work are kept out of search until they have it.

The first two figures are that company’s. The other two are from other clients: one where the work was to be found in Google, and one where it was to be found by AI assistants.

A town's page on Mammoth Roofing's site: dated wind and hail reports logged near Austin, each on its own card.
Each town the company serves has its own page, built on that town’s storm record.

Systems · Mammoth Roofing

+220%

Website enquiries a month

Once the forms and alerts were repaired, website enquiries rose 220% in a month.

Source

The old site’s form records, June and July 2026, against enquiries recorded in Google Analytics, September 2026. A recovery after the forms had failed, counted in two systems.

Build · Mammoth Roofing

−74%

Home page weight

The new home page is 74% lighter than the old one, with almost no layout shift and 58% less time blocked.

Source

Google PageSpeed Insights, mobile, home page: the old site on 7 August against the new site on 21 November 2026, the median of six runs.

SEO and AI search for roofers

For roofers and roofing companies in the UK, and for roofing contractors in the US.

Questions

+000%

Your number here.

We read it the way Google and the AI assistants do. It takes a few seconds.

Find out how well the assistants can read your firm’s site.

We read every brief the same day and reply with a plan.

Start a project