We Checked 100 Houston-Area Websites. Here's What We Found, October 2026.

We ran Google's PageSpeed Insights against each home page on mobile, three times each, and kept the median. The median lab performance score was 62 of 100, and 23 of 100 scored under 50. 97 of 100 took longer than 2.5 seconds to show their main content in that lab test.

Here's the catch: a lab test simulates a slower phone on a slower connection. It isn't what a real visitor on a real phone actually experienced. For that, Google keeps separate real-visitor data, the Chrome User Experience Report, and it exists for only 6 of 100 of these sites. So the more interesting number isn't "97 of 100 are slow in a simulated test." It's this: for 94 of 100 Houston-area sites we checked, nobody, including Google, has real-visitor speed data at all. Whatever these sites' actual visitors are experiencing, it isn't being measured by the tool most of the web relies on.

At 390 pixels wide, roughly a typical phone screen, 98 of 100 sites had a working viewport tag and 90 of 100 had no sideways scrolling. 70 of 100 had a tap-to-call link on the home or contact page, and 64 of 100 passed all three checks together.

The one worth sitting with: 30 of 100 have no tap-to-call link. Not "no phone number," no tap-to-call link; a site can list a number that still isn't a link a thumb can tap. On a phone, that's the difference between a visitor calling you and a visitor giving up.

81 of 100 have a separate contact page; the rest keep contact details on the home page. 70 of 100 have an inquiry form, and 60 of 100 have a form with labelled fields and no script errors on load. 87 of 100 give a visitor some way to reach them, a form or a tap-to-call link.

We never submitted a single one of these forms. Sending 100 businesses a fake inquiry is spam, not research, so we can't tell you whether any given form actually delivers to an inbox, a vendor plugin, or nowhere. What we can tell you is whether a form exists, whether its fields are labelled, and whether it throws a script error the moment the page loads. The 13 of 100 with no form and no tap-to-call link may still list an email address or a plain phone number; we're not calling them unreachable, only harder to reach in the two ways we checked.

We found a Google Business Profile clearly belonging to the business for 83 of 100. "Found" means a name-and-city search turned up a listing we could confirm was theirs; it doesn't mean the other 17 have no listing anywhere, only that we couldn't confirm one.

Of the 83 we found: 79 link to the site, 75 list hours, and 62 show a phone number that matches the one on the site. 15 of those 83 show a phone number that doesn't match. Put together, 60 of 100 have a listing that's found, linked, with hours and a matching phone.

That 15 of 83 is the one we'd act on first if it were our business. A visitor who calls the number on your Google listing instead of the one on your site, because that's the one they found first, and reaches the wrong line, a disconnected one, or nobody, never gets to tell you they tried.

86 of 100 have a robots.txt file at all. Of all 100, none blocks an AI search crawler. None blocks a training-only crawler either, which is a choice some sites make on purpose, not a mistake. 29 of 100 publish an llms.txt file.

That runs against the usual pitch that small-business sites are quietly locking AI search tools out. By the rule these sites actually publish, they aren't. We're reporting robots.txt rules only: a firewall or a bot-blocking service upstream could still stop a crawler, and we didn't pretend to be one to test that.

We drew our 100 from Houston-area businesses and organizations listed in the Pearland, The Woodlands and Tomball chamber directories. Ten internal sampling groups of ten sites each, drawn as every k-th listing from a random start, with chains, franchises, large firms and public bodies removed before the draw. Because every group is exactly ten, the total weights each group equally; it isn't a picture of how the chamber directories themselves are split by industry, and we aren't publishing a breakdown by group.

We ran the checks on October 16, 2026: two pages per site, the home page and the contact page, one visit each, a user agent naming Howard Creative Co., and nothing on a page a site's own robots.txt closed to every agent. Six sites closed themselves to unnamed agents and one refused every visitor; we replaced all seven with the next site in their group and logged the swap.

With 100 sites, each figure here is roughly plus or minus 10 points (estimated, 95% confidence, assuming simple random sampling; our draw was stratified by group, so treat this as a rough guide, not a precise error bar).

What none of this tells you: the lab-versus-field speed split (above), whether a form that exists actually delivers anywhere, whether "listing not found" means no listing exists or just that we couldn't confirm one, what a firewall might block that robots.txt doesn't say, and whatever an automated accessibility scan can't catch on its own. Lab speed is one region and one simulated device profile, not every visitor's experience.

One more thing, said plainly: Howard Creative Co. sells fixes for these kinds of problems, and we may contact the owners of sites we checked here.

No business is named on this page, in a post about it, or anywhere else we publish it. No URLs, logos or screenshots of a real site appear here; any image on this page is a mock-up.