AI visibility for local businesses: what ChatGPT says when someone asks for a plumber

178 sites in our readiness sample had a homepage that answered a plain request. That is a strong starting point, but it does not show whether ChatGPT will recommend a local business when someone asks for a plumber.

That is the practical problem behind local AI visibility. A local business needs to be identifiable, understandable as a service and supported by sources that answer local questions. Maps data and review sites are part of that picture. Your website still matters, but it has a different job from a map profile or a review page.

Local business signals

When someone asks ChatGPT for a plumber, the answer needs more than a name and a phone number. It needs a location, a service category, evidence that the business serves the area and enough context to decide whether the recommendation fits the request. Local answers therefore draw on signals outside your website as well as the pages you control.

Maps data gives an assistant a structured local entity to work with. Review sites add third party evidence about the business and its service. Your site supplies the facts those profiles often cannot explain cleanly: which jobs you take, which areas you cover, how customers contact you, what an emergency call means and what a buyer should expect before booking.

The sources do not all play the same role. A map listing can help establish that a business exists in a place. A review page can support reputation. A service page can explain the match between a buyer's problem and the work you do. If those signals disagree, the assistant has less reason to treat the recommendation as safe.

Our broader citation research shows how little of the answer space belongs to a site's own pages. The scanned product's own sites took 1.7% of recorded citations, while review platforms took 0.8%, as shown in the weekly research on who owns AI answers. That sample covers AI recommendations broadly, not local plumbers, so it is not a local market share. It is a warning against treating your website as the only source that matters.

For a local business, the sensible question is whether the whole set of local signals says the same thing about who you are, where you work and what you do. A plumber's website, map listing and review profile should use the same business name, phone number, service area and core services. If the site says Cambridge while a profile points to another town, or one page says emergency work is available while another says nothing about it, the assistant has conflicting evidence. A homepage can be easy to reach and still be a poor basis for a local recommendation.

Website facts

Your site should make the basic answer easy to extract. Put the business name, service, places served and contact route in ordinary text. Give each important service its own page when the work differs. Keep the wording literal. “Boiler repairs in Cambridge” gives an assistant more usable material than a slogan about dependable solutions.

The AI readiness checks show a split between technical access and commercial clarity. AI crawlers were allowed in robots.txt on 175 of 182 sites. Content was present without JavaScript on 146. A clear H1 passed on 151 sites, and title and description wording passed on 137. These checks matter because a page cannot be understood reliably if the important content is blocked, hidden or vague.

The weaker checks are closer to the decision a buyer is actually making. A plain statement of what the business is, for whom, at what price and against what passed on only 33 of 182 sites. It warned on 96 and failed on 49. A pricing signal passed on 145 sites, and that check counts either a visible price or a link to a pricing page as a pass, so a local business does not need a single published price to satisfy it.

The wording points to a missing sentence rather than another piece of markup. A plumber might need to say which customers are served, which jobs are covered, whether emergency work is available and how estimates work. The exact answer depends on the business, and the check does not tell you what to charge or promise. It tells you that the page should make the choice legible.

Before the change, the page has a large hero line saying “Quality you can trust”, a stock image, three service names in a menu and a phone number in the footer. After the change, the opening text says that the business provides boiler repair and emergency plumbing for homes in Cambridge and nearby areas, explains which jobs it accepts, states how customers request a quote and links to the relevant service pages. The revised page is easier for a buyer, a crawler and an assistant to use because its basic facts appear in the text instead of being left to inference.

Readiness checks

The checks worth fixing first are the ones that affect access and interpretation. Start with robots.txt, rendered content, a clear H1, page titles and descriptions, and a homepage that answers a plain request. The free AI readiness check shows these checks together rather than treating one technical pass as proof of visibility.

Then inspect the checks that describe the business. Organization schema passed on 113 of 182 sites, while 65 failed. That markup can help machines connect a name with an organisation, but it cannot repair a page that never says what the business does or where it operates. Use it as supporting structure, not as a substitute for plain language.

Canonical and Open Graph tags passed on 140 sites, with 38 warnings. A sitemap passed on 153. These are sensible housekeeping checks. They help with page identity and discovery, but they do not answer the buyer's local question by themselves. A tidy sitemap does not tell ChatGPT whether you repair a leaking tap, serve a particular town or take weekend calls.

Questions answered on the page passed on 99 sites and warned on 79. That is more relevant to local work than a site owner may expect. Buyers ask whether a service is available in their area, what it covers, whether it is urgent and how much information they need to provide. Write those answers where the service is described, rather than placing them in a hidden FAQ with no clear link from the page.

Product or SoftwareApplication schema is a poor priority for a plumber. It passed on 102 of 182 sites and failed on 76, a split that may simply reflect how many of these businesses are not software products at all, so the check is unlikely to be the right one to chase. Local service companies should not add product markup simply to collect a pass. Use the checks that describe the business you actually run.

The readiness series makes the pattern plain. Crawler access and homepage reachability are common, while the statement that connects service, customer, price and alternative is much less common. That gap is where a local website can become more useful without turning into a technical project.

Source coverage

A recommendation does not have to cite your own site to be influenced by your local presence. An assistant may use a map record, a review platform, a directory or a local roundup for the recommendation, then use your website to check what the business does. In our readiness sample, only 25 of 182 sites had their own pages cited by an AI answer at all, evidence that your website's job is usually to inform the decision, not to be the source quoted. It may also cite a page you do not control.

The wider citation sample is heavily concentrated in listicles and roundups, which took 68.3% of recorded citations. Competitor product pages took 24.1%. Together they accounted for 92.4%. As noted above, these figures come from product recommendation scans rather than local service queries, so they are not proof that a plumber needs a national roundup. They do show that assistants often cite pages that compare or collect options rather than the provider's own page.

For local businesses, the comparable sources may be local directories, area guides, review pages and service roundups. The exact mix needs measuring. Review platforms accounted for 0.8% of all recorded citations in the broader sample, which is small, but a local recommendation can still depend on reputation signals even when the final answer cites another page.

This changes what you should check. Search for your business in the map systems and review sites your customers use. Check that the name, phone number, location and category agree. Read the service description as a buyer would. If a profile says one thing and your website says another, fix the contradiction before adding another article.

You should also check the pages assistants actually cite. The weekly research on who owns AI answers is useful for seeing the page types that appear in the broader sample. For your own business, a visibility check should record the question, the engine, the business names mentioned, the source pages shown and the local details used to justify the answer.

Low-priority checks

Do not begin with every readiness check. A local business does not need to chase a perfect technical score before it can make its service clearer. Product schema, llms.txt and elaborate comparison pages may be irrelevant to the immediate problem, depending on the business and the question. The sample found llms.txt on 110 of 182 sites, but that does not establish that adding the file makes a local recommendation more likely.

Comparison or alternatives pages are a good example of a low-priority check for a local service business. The check passed on 91 of 182 sites, failed on 41 and warned on 50. A plumber usually needs to explain its own services and coverage, not publish a comparison of plumbers. That check may fit a marketplace or a business selling against named alternatives, but it is a poor first task when the core local facts are still unclear.

A visible price is not a complete commercial answer. A fixed price may be impossible for a repair service, and a warning on the pricing check does not mean the site is unusable. Say how quotes work, what information the customer should provide and whether there are different service conditions. That is more honest and more useful than placing an invented starting price on every page.

Organization schema can support local identity, but it cannot settle it on its own. Keep the name, address, phone number, service area and opening information consistent in the visible page text and in the profiles customers use. An address on the site that differs from the map profile is a simple example of the problem. Structured data can reinforce a fact, but it cannot create one that the rest of the web contradicts.

Generic “best plumbers” articles are not automatically the answer just because roundups dominate the broader citation sample. A local business may benefit more from a clear service page, an accurate profile and useful answers to local buyer questions. Page type is a clue rather than a command. Finally, a readiness pass is not a recommendation: the checks show whether a site exposes useful signals, not whether the business is good, available, affordable or right for a particular customer.

We measured samples, not the whole web. The readiness results come from 182 audited sites, and the citation results come from recorded scans. We do not know how ChatGPT handles every local query, map source or review platform. The caveat that matters most is simple: a technically ready website can still be absent from a local answer if the surrounding local evidence does not support it.