Why AI Search Ignores Your Plumbing Website

Usually for 3 mechanical reasons, not a content one. As of Vercel’s December 2024 crawler study, no major AI crawler executes JavaScript, so script-injected content is invisible. Your name, hours and service area disagree across profiles, so nothing is verifiable. And your answers run 200 words when an engine can only lift 40 to 60. Fix access first, then extraction, then consistency.

Picture the moment the work is actually won. It is 2am, a supply line has let go under a kitchen sink, and a homeowner standing in 1 inch of water is not opening 5 tabs to compare quotes. They ask an assistant who to call, it names 3 plumbers, and one of them gets a job worth several hundred dollars before anyone has looked at a website.

Plumbing is the trade where this hurts most, because emergency intent does not shop, it picks. The gap between being named and not being named is the entire transaction. There is no second-place consolation, no retargeting, no nurture sequence. The list has 3 slots.

What makes this genuinely frustrating for owners is that the sites being skipped are often fine. They load, they rank respectably in classic Google Search, they have real reviews and a decent gallery of work. The failure is not quality. It is that an assistant is running a different set of checks than Google is, and a site can pass one while failing the other completely.

Why does AI search ignore your plumbing website?

Because it cannot read, verify or lift what you published. The 3 failures compound: content injected by JavaScript never renders for the crawler, details that disagree across profiles fail verification, and answers buried 200 words deep cannot be extracted. Emergency intent then decides in seconds, without you.

The 3 checks an assistant runs first

These happen in sequence, and each one gates the next. A site that fails the first never reaches the third, which is why owners who invest in better copy sometimes see nothing change at all.

• Access. Is the crawler allowed in robots.txt, and is the content present in the HTML that gets served?
• Verification. Do the business details agree across the sources the engine can independently check?
• Extraction. Is there a self-contained passage short enough to lift into a generated answer?

The order matters more than any individual fix. Rewriting your service pages to be beautifully quotable does nothing if the crawler was disallowed 2 years ago and nobody revisited the file.

Does an assistant read your site the way Google does?

No, and the gap is the whole problem. Googlebot and Applebot render JavaScript. According to Vercel’s December 2024 study, GPTBot, ClaudeBot, OAI-SearchBot, ChatGPT-User and PerplexityBot do not. They fetch script files but never execute them, so client-side content is simply absent from what they read.

Inside a single crawler fetch

The measured behaviour is specific rather than theoretical. Vercel found ChatGPT’s crawlers requesting JavaScript files in 11.50% of requests and Claude’s in 23.84%, while executing none of them. At the volumes involved, GPTBot alone accounted for roughly 569 million requests a month, with ClaudeBot near 370 million and PerplexityBot at 24.4 million.

For a plumbing site the practical consequence is narrow but expensive. If your service list, service-area map, pricing table or reviews widget is populated after page load by a script or a third-party embed, an assistant sees an empty shell where your offer should be. The page is not penalised. It is blank.

This is also why the usual diagnostic misleads. Viewing the page in a browser shows everything working, because your browser runs the scripts. The honest test is to fetch the page the way a crawler does, or simply disable JavaScript and reload. Whatever survives is what an assistant has to work with. The AI visibility work DaxReach does starts here, because every later fix is contingent on it.

How does a plumber become verifiable to an assistant?

By making the same 5 facts agree everywhere an engine can find them: legal name, phone, service area, hours and licence status. Assistants cross-reference before naming anyone, because recommending a plumber who does not answer is a failure they are tuned to avoid. Contradictions are what disqualify you.

The signals that carry the weight

Verification is unglamorous and it is where most of the recoverable ground sits for a trade business. An engine has no way to assess whether you are good at soldering copper, so it leans on what it can corroborate independently:

• Google Business Profile — correct primary category, real service area, accurate hours, and answered questions.
• Name, address and phone consistency — identical formatting across every directory, not merely similar.
• Licence and insurance status — stated in crawlable text on your own site, not buried in a PDF.
• Recent reviews — a steady flow matters more than a large historic total, because recency reads as still operating.
• Response expectations — plainly stated hours and callback windows for emergency work.

None of this is persuasion, and that is the point. Letting a stranger into your home under stress makes verification the trust gate, so the signals an assistant can confirm are the same ones a nervous homeowner is looking for at 2am.

Which pages win the planned work?

The ones that answer research questions in text. Emergency calls are reflexive, but water heaters, repipes and remodel rough-ins get researched for days, and that research increasingly runs through an assistant. A page answering the sizing and cost questions is what gets quoted mid-decision, when the higher-margin job is still open.

Two funnels, two sets of pages

Most plumbing sites are built for exactly 1 of the 2 funnels and then wonder why the job mix never improves. Emergency work is instant, urgent and relatively price-insensitive. Planned work is compared carefully over days or weeks, and it carries the better margin.

The planned funnel is answered with content, and these are the questions it has to actually address:

• Repair or replace, and the age at which replacing stops being optional.
• Tank versus tankless, including what the install genuinely involves in an older home.
• Sizing and recovery rate for a household of a given size.
• Repipe scope and what happens to the walls.
• Permit and inspection expectations for the work.

Answer those honestly, in extractable text, and you get cited during the research phase. Skip them and you have optimised for the 2am call while leaving the profitable work to whoever bothered.

Should you add schema markup first?

No. Schema is worth adding and it is nearly free, but the evidence says it does not buy citations. Ahrefs tracked 1,885 pages that added structured data against roughly 4,000 matched controls and found AI citations slipped 4.6%. Access and extraction move the outcome; markup describes what is already readable.

The sequence that respects the mechanics

Google states that structured data is not required for its AI features, and DaxReach’s own position on schema and AI citations says the same thing at greater length. Structured data is a label describing the box. If the box is empty because a script never ran, a better label changes nothing.

The crawler documentation is public and worth reading directly rather than through a vendor’s summary: OpenAI publishes its bots, as does Perplexity, and Google documents its own crawlers including Google-Extended.

Emergency plumbing marketing options compared

Every route below genuinely works for something. The honest question is which constraint you are actually under this month.

The row worth reading twice is speed. If the trucks are empty next week, ads and marketplaces are the honest answer, and no amount of AI-visibility work will fill that gap in time. What the organic route buys is the position where the call arrives without a per-lead fee attached, which is a different problem on a different timescale.

The first 3 fixes, in order

If the whole teardown collapses into a short list, it is this:

1. Disable JavaScript and reload your own site. Whatever disappears is invisible to every major AI crawler. Move it into the served HTML.
2. Reconcile the 5 core facts across your Google Business Profile and every directory carrying your name, until they match character for character.
3. Rewrite 4 to 6 buyer questions as headings, each followed by a self-contained answer of roughly 40 to 60 words, then continue at length under a sub-heading.

Then verify rather than assume, by checking whether AI assistants actually mention your business before and after. If the terminology in any of this is unfamiliar, the glossary defines it, and the comparison page sets out where DaxReach fits against the alternatives above.

Being ignored by AI search is rarely a verdict on the plumbing. It is usually a site that was built for a browser to render and a crawler to skip.