The details below are anonymised and lightly composited, as every teardown here is. The business is real in the sense that I've seen this exact shape a dozen times; the name and numbers are not any one company's.

The business

Call it Copper State Comfort Heating & Air. Mesa, Arizona. Founded 2004, family-owned, four trucks, a 4.8 average across a few hundred Google reviews. The owner came to Duly Noted because a customer had mentioned, in passing, that she'd "asked ChatGPT" before calling and Copper State hadn't been on the list. He wanted to know if that was true.

It was.

The report

Overall 38, a D. AI visibility 20, an F. Site readiness 66, a B. If you've read the note on scoring you'll recognise the shape: a site that's mostly fine, attached to a business the assistants barely know.

Across the twelve cells — four assistants, three questions about HVAC repair in Mesa — Copper State was named twice and linked once. ChatGPT named it third on one list. Perplexity linked it once, citing an Angi profile rather than the company's own site. Claude and Gemini never mentioned it. Two competitors appeared in nine and seven cells respectively. Three of the domains the assistants cited most were directories: Angi, HomeAdvisor, and, interestingly, the Arizona Registrar of Contractors, which the assistants were using to confirm licences.

What the site was doing wrong

The readiness audit passed eleven checks and flagged eight. Two were failing.

No LocalBusiness schema. The site had schema — a WebSite block and breadcrumbs from the SEO plugin — but nothing that said "this is an HVAC business at this address." To a crawler it was a website about heating and cooling, not a company that does it.

No meta description. The homepage had none, so every retrieval system that uses the description as its page summary had to guess from the first paragraph, which was a slogan.

Six items needed work. The title tag was "Home - Copper State Comfort," twenty-seven characters, with no trade and no city. Server response was 1.8 seconds, slow enough that some retrieval fetches would time out before reading. The homepage was about four hundred words, most of them adjectives. There was a section headed "Common Questions" with three questions and no FAQ markup. No llms.txt. And the schema that did exist described the site rather than the business.

The things it did right matter too: HTTPS, one clean H1, canonical tags, name-address-phone in the footer, an About page, a Contact page with a map embed, robots.txt that didn't block anyone. Twenty-two years of being a real business showed. The site simply hadn't been asked to explain itself to a machine.

Why "mostly fine" still scores an F

Here's the part owners find hardest. Nothing about this site was broken. A human visitor found the phone number, saw the reviews, called. The problem is that the assistants aren't visitors. They're readers looking for a page that answers "who does HVAC repair in Mesa" in a form they can quote, and this site never said those words in that order anywhere a machine would weight them. The title didn't. The schema didn't. The description didn't exist. The content was a brochure.

The competitor that appeared in nine cells wasn't a better contractor. Its site had a title that said "AC Repair Mesa AZ," a LocalBusiness block with a sameAs link to its Google profile, a service page per neighbourhood, and an FAQ that listed prices. It was legible. Copper State was not.

The fix list, in order

This is the action plan from the report, sequenced by weight.

Add HVACBusiness schema with the full address, phone, service area, hours and sameAs links to the Google Business Profile, Angi and the ROC licence page. Half a day with a plugin. This is the single change that moves an absent business to a named one, because it's the first time the site states the entity.

Rewrite the title tag and write a meta description. Ten minutes. "Copper State Comfort — AC Repair & HVAC Service in Mesa, AZ" and a 150-character description that says what, where, since when.

Turn "Common Questions" into a real FAQ. Eight questions from the phone log — cost of a service call, same-day availability, what areas, tune-up pricing, repair-or-replace — with forty-to-eighty-word answers that commit to numbers. Mark it up.

Add three hundred words of substance to the homepage: services with prices where possible, service area by name, credentials, how the process works. Not adjectives.

Fix the response time. In this case a caching plugin and image compression got it under a second.

Publish llms.txt. Ten minutes, and the exercise produced the description paragraph that went into the schema and the meta.

What we told the owner to expect

Weeks, not days. The crawlers were never blocked, so re-indexing is fast, but the assistants' sense of which businesses belong to "HVAC in Mesa" is built from many sources and updates on their schedule. The realistic target for the first re-run was moving from named-in-two to named-in-five or six, with the company's own site linked at least once instead of the Angi profile. The second target, over a quarter, was displacing one of the two dominant competitors in at least one assistant's list.

Both of those are measurable, which is the point of running the report monthly rather than once.

The lesson

If your site is old, honest and decent-looking and you're absent from the answers, don't assume you need a new website. You probably need a title tag, a schema block and a FAQ page with prices in it. That's a three-week fix, not a rebuild — and it's the reason the Fix Sprint is a flat price.