What a top ten HVAC website actually looks like, measured across twelve markets

Ask around about what it takes to rank a home services website and you will be told, confidently and for free, that you need more pages and a busy blog. We wanted to know whether the sites that already rank actually look like that, so we measured them.

On September 1, 2026 we captured the Google top ten organic results for “HVAC company” in twelve US metro areas: Phoenix, Dallas, Atlanta, Charlotte, Columbus, Denver, Tampa, Nashville, Kansas City, Sacramento, Indianapolis and Las Vegas. That produced 105 results. After screening and deduplication we measured 67 distinct companies, reading each one’s published sitemap and homepage directly.

Here is what the data says, including the part where it refuses to say what we expected.

A third of the top ten is not what you searched for

Of the 105 results captured, 35 were not HVAC companies. Thirty-two were directories, marketplaces or social platforms: Yelp appeared ten times, Reddit eight, a construction bidding marketplace three, plus Thumbtack, Angi and HomeAdvisor. Three more were equipment manufacturers rather than local operators.

That is the first finding and it is not a footnote. If you are a contractor in one of these markets, roughly a third of page one is occupied by companies that will never install a furnace. The competitive set you are actually fighting for position with is smaller than the page looks, and part of your visibility problem is a lead aggregator that ranks for your category and then sells your name back to you.

The median top ten HVAC site carries 378 pages

Across the 57 companies with a readable sitemap:

• Smallest: 13 pages
• 25th percentile: 149 pages
• Median: 378 pages
• 75th percentile: 682 pages
• 90th percentile: 1,038 pages
• Largest: 10,819 pages

The mean is 793, which is roughly double the median. That gap is the whole reason we report the median. A handful of very large sites drag the average up to a number that describes almost none of the companies in the sample. If someone tells you the average top ten HVAC site has 793 pages, they are technically quoting the arithmetic and functionally describing nobody.

Ten of the 67 companies had no readable sitemap at all. They are reported as missing rather than counted as zero, because we do not know how many pages they have. We know we could not see them.

Blog volume splits the field in half

Counting only URLs published under a blog, news, article or resources path:

• 24 of 57 companies have two or fewer such pages, which in practice means no blog program
• The median company has 43
• The 75th percentile has 194
• The largest has 1,919

So half the field is running something that looks like a content operation and half is not, and both halves are on page one. Whatever is separating these companies from the ones on page two, it is not that all of them decided to blog.

What does not correlate

This is the part we expected to go differently.

Within this sample we found no statistically significant relationship between search position and any of the size measures:

• Position and total page count: rho -0.26, n=57, not significant
• Position and blog page count: rho -0.25, n=57, not significant
• Position and homepage word count: rho 0.03, n=62, not significant

Comparing the top of the page to the bottom directly tells the same story. Companies ranking first through third had a median of 305 pages. Companies ranking eighth through tenth had a median of 207. The bigger sites are slightly ahead, and the difference is well inside what a sample this size produces by chance.

Two cautions on how to read that, both of which matter more than the numbers.

First, not significant is not the same as no effect. With 57 companies we can only detect fairly strong relationships. A real but modest effect would be invisible at this sample size, and we cannot rule one out.

Second, and more important, everything here was measured on sites that already rank in the top ten. Comparing winners to other winners cannot tell you what separates winners from everyone else. If page count is a threshold you have to clear to compete at all, this study would not see it, because every company in the sample already cleared it. The honest claim is narrow: among sites that already rank, size does not predict where in the top ten they land.

The one strong relationship we did find is the uninteresting one. Page count and blog page count move together at rho 0.58, which is significant and also obvious, since blog posts are pages. That relationship is worth naming precisely because it is how the folk wisdom gets manufactured. Sites with blogs are bigger. Bigger sites are assumed to rank better. The inference gets made without anyone checking the step in the middle.

What the field has actually standardized on

Structured data and trust signals were more consistent than size, across the 62 companies whose homepage we could read:

• 87 percent publish some JSON-LD structured data
• 44 percent identify themselves with LocalBusiness schema
• 31 percent publish review or rating schema
• 84 percent link to careers or hiring from the homepage
• 79 percent mention financing
• 81 percent link to a blog or resources section

The gap between the first two lines is the interesting one. Nearly everyone is emitting structured data, usually because their platform does it automatically, and fewer than half are using it to state the single fact that matters most for a local contractor, which is that they are a local business serving a place. That is a gap you can close in an afternoon, and unlike page count it has a defined right answer.

The finding that changed when we fixed the sample

We ran this study twice.

The first run reported a moderate, statistically significant relationship between page count and position, rho -0.35. That would have made a better headline than the one we ended up with.

It was wrong. The sample contained an HVAC industry directory carrying 54,097 pages and ranking second in one market. It is not a contractor, it does not compete for installations, and its presence pulled a large page count into a top position and dragged the correlation with it. Removing that single row moved the result from significant to not significant.

One contaminated row out of 57 changed the answer to the research question. Not the confidence interval, the answer.

We are reporting that because it is the most useful thing in this study. Every number above depends on a screening decision made before any measurement happened, about what counts as an HVAC company. Those decisions are usually invisible in published research, and they are frequently where the finding actually comes from. Ours are written down: directories, marketplaces, social platforms, manufacturers and industry guides were excluded, every excluded row is retained in the dataset with its reason, and the count is published above.

What we measured, and what we did not

Search results were captured once, on September 1, 2026, with one location setting per market. Rankings vary by device, personalization and time, so this is one photograph, not a trend.

Page counts come from published sitemaps, which is what the site itself claims to have. Blog volume counts URLs under a blog, news, article or resources path, so a company publishing under a different structure is undercounted. Recency uses sitemap lastmod values, which record modification rather than strictly publication.

Position is position within one query. It is not overall visibility, traffic, or revenue, and nothing here should be read as a claim about any of those.

We did not measure backlinks, Google Business Profile completeness, review counts, page speed or ad spend. Any of those could carry more of the explanation than anything we looked at.

Why this is the honest version

The version of this article that performs better writes itself: measure some sites, find a correlation, publish “top ranking HVAC sites have twice as many pages,” and let readers conclude that adding pages causes rankings.

We had that article. It survived until we checked the sample.

The reason to work this way is not fastidiousness. It is that a marketing decision made on a finding that dissolves under scrutiny costs real money, and the person who made it has no way to know it dissolved. Publishing the screening rules, the missing data counts and the sample size is what makes a finding checkable by someone who did not run it.

If you want the underlying figures for your own market, or you want to know what this method says about your category rather than HVAC, that is a conversation worth having.