What an AI Engine Reads on a Contractor Website

Open a contractor website, turn the images off, and read what is left. That is close to what a machine gets. On most construction sites what remains is a company name, a menu, a phone number and three paragraphs of general language about quality and communication. An engine asked to name a commercial concrete contractor in a specific county has nothing in that to work with, so it names a company that gave it something.

What does an AI engine see when it reads a contractor website?

An AI engine sees text and structured data. Headings, body copy, link text, alt attributes and schema markup are readable. Photographs, background video, slider animations and text baked into images are not. A contractor site carrying its proof in pictures is invisible on the surfaces that answer questions.

This is why portfolio pages underperform so badly. A jobsite album is real evidence of capability and it carries no meaning without written scope beside it. The same applies to capability statements published as downloadable PDFs that never appear as page content.

What is an entity block and why does a contractor need one?

An entity block is the plain statement of what a company is, where it is based and what it does, repeated in identical wording across every page. It appears in visible prose in the first 150 words and again in structured data. Engines use it to decide which company a page describes.

Schema.org, the shared vocabulary search engines read structured data from, publishes a LocalBusiness type with a HomeAndConstructionBusiness subtype built for this purpose, alongside Organization and Service types. Google is blunt about the limits here, stating in its guidance on AI features that no special schema.org markup is needed to appear in AI Overviews or AI Mode. What markup does is remove ambiguity about the category, and it has to match the visible text on the page.

The markup has to agree with the visible page. Name, address and phone in the schema, the footer, the contact page and every directory listing should match character for character. A suite number in one place and not another is enough to split one company into two records, and a split record gets cited less than either half would have.

Which facts are usually missing from contractor websites?

Most contractor sites carry two or three of the facts an engine needs. The commonly missing ones are the trades performed by name, the counties served, typical project size, license and registration numbers, whether the company self-performs or subs out, residential versus commercial mix, and bonding and insurance limits.

Read a homepage and count. Every absent fact is a question the engine answers by leaving the company out of the response. This is the single largest gap on contractor sites and it costs nothing to close, because the information already exists inside the business.

Naming people belongs on the same list. A construction background is a real differentiator and an unnamed one is unverifiable. Bios with names, roles and years in the trade give an engine something to attach the claim to. That is also where owned content stops being enough, since verifiable people are what third parties end up quoting.

Do registration numbers belong on a contractor website?

In several states a registration number is required in advertising, and a website is advertising. Pennsylvania contractors performing home improvement work must register under the Home Improvement Consumer Protection Act and display the registration number clearly in all advertisements, contracts, estimates and proposals.

The Pennsylvania Office of Attorney General states the requirement plainly in its HICPA guidance, updated 25 June 2026: the number must use the PA prefix, must appear in advertisements, contracts, estimates and proposals, and must be clearly and conspicuously displayed. The act applies to work on private residences, not commercial property, and registration must be renewed every two years.

Compliance is the first reason to publish the number. The second reason is that a registration number is a verifiable identifier, and verifiable identifiers are exactly what an engine can attach to a company with confidence. A contractor who publishes a checkable number is easier to name than one who does not.

What parts of a contractor website can a machine not use?

Machines cannot read photo galleries without captions, capability statements locked inside PDFs, contact details set as images, video without transcripts, or testimonial sliders built in scripts that load after the page. Each of those is a place where a contractor’s strongest proof stops at the browser.

Video is worth calling out because it is the easiest recovery. A construction company with a YouTube channel is already ahead of most of its trade. The transcript is the part a machine reads, so publishing it alongside the embed converts a video asset into text an engine can quote.

Testimonial sections deserve the same attention. An empty block with a heading promising client feedback is worse than no block at all. It reads as an abandoned page to a human and returns nothing to a machine.

How should a contractor write a page so an engine can quote it?

Write short, self-contained answers. A passage that works when it is lifted out of the page and read alone is the passage that gets used. Name the company, the trade and the region inside the answer, open with the answer rather than a preamble, and refer to nothing above or below on the page.

That rules out a common habit. “As mentioned above, our team handles this” is unusable, because the engine takes the passage and not the context around it. So is “we” without a nearby company name.

Project pages follow the same discipline. An entry with six photos and a company name is a dead end. The same entry with the scope, the square footage, the system installed, the schedule and the problem solved is a page an engine can quote and an owner can evaluate.

What order should this work be done in?

Access first, accuracy second, depth third. Confirm crawlers are allowed and pages are indexed. Correct the entity facts and the schema so every reference to the company agrees. Then write the answers. Reversing that order produces good writing on pages nothing can reach.

The full sequence for a contractor runs longer. Unblock crawlers, submit the sitemap to Bing as well as Google, resolve duplicate pages and redirect chains, claim and complete the business listing, add Organization and LocalBusiness schema, write the depth pages, publish video with transcripts, then pursue trade coverage. FAQ markup no longer earns a rich result in Google, which dropped that feature in May 2026, so add it only where it describes questions a visitor can see.

The technical steps are cheap and fast. Everything downstream depends on them, and a contractor who skips them is paying for content on pages that no engine can see.

Frequently Asked Questions About Contractor Websites and AI Engines