13 First-Page HVAC Companies. AI Named Them in 80 of 260 Answers.

In August we tested thirteen roofing companies and found that twelve never had their own website linked in an AI answer. The fair objection was that the sample was mixed: some ranked well on Google, some barely ranked at all. So for the second study we took that variable out. Every HVAC company here already holds Google's first page for "hvac company [city]" in its own metro. If anyone should be getting recommended by ChatGPT, Gemini, Claude and Perplexity, it is these thirteen.

They did get in. Twelve of the thirteen had their own site linked in at least one answer. But the tested company was named in 80 of 260 answers, 31 percent. The other 180 answers recommended someone else.

13

HVAC companies, one per metro, each on Google's first page for its city

260

answers read: thirteen companies, five prompts each, four assistants

12 of 13

companies had their own website linked in at least one answer

80 of 260

answers named the tested company, 31 percent

196

citations to the single most cited source, an HVAC directory

As in August, this is not a judgement on anyone's work. It is a measurement of what four assistants do with a local HVAC question, and of what they build the answer from.

How we picked the 13 HVAC companies and what we asked

We pulled Google's organic results for "hvac company [city]" in thirteen US metros from Ahrefs SERP data in October 2026 and took the first HVAC-first company site in each. Not a directory listing, not a plumber who also does HVAC: the first company whose main trade is heating and cooling. One company per metro: Boston, Atlanta, Seattle, Minneapolis, Columbus, Charlotte, Tampa, Dallas, Phoenix, Chicago, Nashville, Denver and Philadelphia.

Five prompts per company, written by the testing tool for that company's market and services. Every set was built from the same five shapes:

1. 01Who are the top HVAC contractors in [metro] in 2026? The plain list question.
2. 02Which HVAC company in [metro] do you recommend for [emergency repair]? Written for whatever the company sells: emergency furnace repair, emergency AC repair.
3. 03Which heating and cooling company in [metro] offers the best value?
4. 04Which HVAC contractors in [metro] have the best reviews?
5. 05Who does the best [specific job] in [metro]? or Where can I find [specific job] in [metro]? Ductless mini-splits, commercial rooftop units, cold-climate heat pumps.

Each prompt went to ChatGPT, Gemini, Claude and Perplexity through the same visibility-testing connector we used in August: fresh sessions, no account history, no location set. The models are the smaller, faster tier of each assistant, not the flagship consumer apps, so the wording you see in the app will differ. Twenty answers per company, 260 in total, collected on October 3, 2026. For reporting, the tool groups the prompts into three families: best in area, how to choose, where to find.

Two numbers per company. Linked, of 5: in how many of the five prompts did at least one assistant put a link to the company's own website in its answer. Named, of 20: in how many of the twenty answers did the company's name appear at all, with or without a link. August scored links only; here the gap between the two numbers is the story.

The results, all 13 companies

Sorted by prompts linked, then answers named. Mention counts in the last column are across that company's twenty answers.

The table is available as a spreadsheet: download the study data as CSV, 13 rows, one per company, with the metro, both scores and the names each assistant gave instead. Every one of the 260 answers is published too; the data appendix at the end explains the file.

What the 260 answers say

1. Companies on Google's first page do get into AI answers

Twelve of the thirteen had their own website linked in at least one of the five prompts. Three were linked in all five: Atlantic Heating & Air Conditioning in Boston, Reliable Heating & Air in Atlanta and Greenwood Heating and Home Services in Seattle. Holding Google's first page clearly goes with being something the assistants can find and cite.

2. Being on the first page is not the same as being the answer

The tested company was named in 80 of 260 answers. The other 180 handed the buyer a different name. Even the Boston company, linked in all five prompts, was named in 11 of 20 answers. Standard Heating & Air Conditioning in Minneapolis had the best named count in the set, 15 of 20, and the single most cited source across its twenty answers was still a competitor's website, goblueox.com, at 19 citations.

Denver is the sharpest case. UniColorado's site carries 3,026 organic keywords and 980 Google AI Overview references in the tool's domain data. It was named in 1 of 20 answers. A first-page ranking gets a company into the room. It does not get it the sentence.

3. The four assistants do not agree

Gemini named the tested company in 31 of its 65 answers. Perplexity named it in 13. We can see what each assistant cited but not how it weighed it, so we will not guess at the cause. The practical point: a company can be the default answer in one assistant and absent from another, so test all four.

4. What the assistants read: directories, review aggregators, and other contractors' sites

The most cited single source in the study is serviceagent.ai, an HVAC directory, at 196 citations. ConsumerAffairs is second at 132. Another HVAC-only directory, hvacservice.io, took 86. Reddit threads were cited 45 times. Three lead marketplaces, which we do not name, were cited 113 times between them. The one company website in the top thirteen, goblueox.com, is Blue Ox Heating & Air, the name given most often instead of the tested Minneapolis company.

By type: company websites and other pages 1,426 citations, spread thin across many domains and mostly other HVAC companies' sites; review and directory sites 310; HVAC-only directories 284; lead marketplaces 113; Reddit 45; manufacturer dealer locators 7.

Those seven dealer-locator citations matter more than the count suggests. In Denver, the four names given most instead of the tested company were Carrier (9), Trane (6), Daikin (4) and Mitsubishi Electric (4). Not one local contractor in the top four. Part of that is the prompts: the tool's five questions for the Denver company asked about heating replacement and cold-climate heat pumps without naming the city, and when a question has no place in it the assistants reach for dealer networks. Across the whole study Carrier was the most frequent name on the named-instead list, 13 mentions, with Trane at 11 and Lennox at 9. Manufacturers that install nothing were the answer to a contractor question.

5. The one company named in none of 20 is the smallest site in the set

Downtown Mechanical in Philadelphia was named in none of 20 answers and linked in none of 5 prompts. It also has the smallest footprint in the set: 106 organic keywords and 2 Google AI Overview references in the tool's domain data, against 2,263 keywords and 398 references for the Minneapolis company named 15 times. Its twenty answers were built from serviceagent.ai (16 citations), consumeraffairs.com (12) and reddit.com (6). Being first on one query is not a footprint; the assistants read dozens of sources per company, and a small site shows up on few of them.

6. Which questions a first-page company wins

The tool's three prompt families, and how often the tested company was named in each: best in area, 42 of 124 answers; how to choose, 28 of 100; where to find a specific service, 10 of 36. The where-to-find questions are the closest to a buyer about to call, and they had the lowest named rate. Those are the questions a page about one job in one place answers.

How this compares with the roofing study

Carefully. The August roofing sample was not picked from Google's first page; it mixed companies at every level of search footprint and scored links only. So 1 of 13 linked in August against 12 of 13 in October is not the same measurement taken twice. What the two studies agree on is the wiring. In both, the answers were built from directories, review aggregators, a handful of marketplaces and other contractors' pages, and in both a manufacturer that installs nothing was among the most named brands: GAF for roofing, Carrier for HVAC.

What this means for an HVAC company

If you hold the first page for your city, the assistants can find you. The work is making sure that when they build a list for your metro, yours is one of the names. In the order we would do it:

1. 01Run the test on your own company first. Five prompts, four assistants, twenty answers. Write down who is named instead and which sources the answers were built from. We run it free and send the raw answers.
2. 02Be on the directories the assistants actually read, with the same name, address, phone and service list as your website. The domains in the chart above are the list. An old phone number on a directory is what gets quoted.
3. 03Reviews on more than one platform. ConsumerAffairs alone was cited 132 times. Volume on two or three platforms is what the answers reach for when they justify a pick.
4. 04A page per job and per town. The where-to-find prompts had the lowest named rate, 10 of 36. A page about ductless installation in one suburb, with the brands you install and the towns you cover, is something an assistant can hand to a buyer as the answer. It is the same work as HVAC SEO in ordinary search.
5. 05A complete Google Business Profile. Categories right, every service listed, hours and phone matching the site, photos current. It is the record behind the first-page ranking you already hold, and the directories copy from it.
6. 06Keep your own pages citable. A FAQ written as questions with short direct answers, specifics instead of adjectives, and one consistent name, address and phone everywhere.

To see where your company lands, run the AI visibility test on your own business, free, and we will send the twenty raw answers. If you want it re-run every month with the fixes done for you, that is the AI search visibility work; the pages per job and per town are HVAC SEO. The August study of thirteen roofing companies is at AI visibility study: roofing.

Data appendix

Two files sit behind this article. The summary has 13 rows, one per company. The answer-level file has 260 rows, one per answer, for all thirteen companies. Every number in the article can be recomputed from it.

How it was collected

• When: October 3, 2026. The tool recorded the date for these runs, not the time.
• How: a third-party visibility-testing tool, called through its API. It writes the prompts, sends each one to the four assistants with no account history and no location, and records what comes back.
• Models, as recorded: ChatGPT gpt-4o-mini, Perplexity sonar, Gemini gemini-3.1-flash-lite, Claude claude-haiku-4-5. These are the smaller API models, not necessarily what the consumer apps answer with.
• Prompts: written by the tool from each company's website, five per company, 65 in all. Most ask for the best or most recommended company in the metro, ten ask where to get a service, and two ask for an alternative to a named company. The tool tags each one mofu or bofu (comparing, ready to hire); the tag is in the file.
• One run per company. Nothing was repeated, so the file is one day's answers, not an average.

What the columns mean

• own_site_linked: the tool's flag that the answer put the company's own website in front of the reader. A prompt counts toward linked, of 5, if any of its four answers has it.
• company_named: the company appears in the tool's list of brands named in that answer, link or not. This is named, of 20.
• cited_domains and cited_urls: the sources the tool recorded for the answer, the pages the assistant drew on. A source is not a click; nothing here measures whether a homeowner followed anything.
• answer_excerpt: the first 400 or so characters of the answer. The tool stores only that much, so the full answers were never kept.

Checks and edits

• Recomputed from the answer file, all thirteen companies match the published scores on both counts, and the total is 80 of 260 answers naming the tested company.
• The three lead marketplaces the article reports as one line appear as [lead marketplace] in the sources, the links and the answer text. Two pages from marketing firms appear as [marketing firm, withheld]; we don't name other agencies. Nothing else is edited.

Both files are free to use under CC BY 4.0, with credit to Riptide Growth.