Most "small business website statistics" pages you'll find recycle survey numbers that are years old. Nobody goes back and checks.
So we did. In September 2026 we took every Orange County, California business that lists a website on the public map data (dentists, lawyers, contractors, auto shops, salons, clinics and more), removed the chains, and tested all 662 of them the same way: does the local business website load, can a phone visitor call or send a request, does Google get the basic signals, how fast is it on mobile, and what does it tell AI crawlers?
Some results were worse than we expected. One was the opposite of what the AI-search crowd keeps saying. All the numbers are below, with the method, so you can check our work.
Key findings:
Want to know which of these apply to your site? Send us the link. We'll run the same checks on your website and tell you the three fixes that matter, free. Book a free website and SEO audit.
How we built the study, in one paragraph
We pulled every business in Orange County, California from OpenStreetMap that has a website listed and belongs to a local service category, then dropped chains (anything tagged with a brand, or sharing one domain across four or more locations). That left 662 unique websites.
For each one we fetched the homepage like a phone browser would, read the HTML, checked robots.txt, llms.txt and the sitemap, sent the same request with AI crawler user agents, re-checked the pages whose raw HTML looked empty in a real headless Chrome, and ran Google's PageSpeed Insights mobile test on every homepage that loaded. The full method and its limits are at the end of this article.
Finding 1: 15% of listed local business websites don't work
This one surprised us the most. Before we tested anything else, 101 of the 662 websites listed for these businesses were simply broken.
| What happened when we opened the link | Websites | Share of 662 |
|---|---|---|
| Loads normally | 489 | 74% |
| Firewall or bot wall, could not be measured | 72 | 11% |
| Domain no longer exists | 54 | 8% |
| Server returns an error page (404, 5xx) | 31 | 5% |
| Broken SSL certificate, timeout or parked domain | 16 | 2% |
Some of these businesses have closed, and that's part of the story. But plenty are still open, still answering the phone and still listed on maps, directories and review sites with a link that goes nowhere. Every one of those listings sends a customer to an error page.
If you haven't clicked the website link on your own Google Business Profile this year, do it today. It takes ten seconds.
Finding 2: 26% of homepages have no tap-to-call link
A lot of the people who land on a local business website are on a phone, and many of them want one thing: to call. We checked whether the homepage has a real `tel:` link, the kind that dials when you tap it.
- 26% of live homepages have no tap-to-call link.
- 12% show a phone number as plain text, so the visitor has to memorize it or copy it by hand.
- 15% show no phone number on the homepage at all.
That's the cheapest fix in this whole study. A developer can add it in minutes, and on most builders you can do it yourself.
"A phone number you can't tap is a phone number most mobile visitors won't dial."
Finding 3: 25% give visitors no form and no booking
Not everyone wants to call. Some visitors are at work, some want a quote in writing, some are browsing at 11 p.m. We looked for any way to send a request from the homepage: a contact form, a quote form, or an online booking widget.
25% had none. On the other side, 38% offer online booking of some kind; 84% of those are health, dental, beauty or fitness businesses using scheduling tools.
Put findings 2 and 3 together and 12% of live homepages have neither a call link nor a form. The visitor's only way to reach them is to go looking.
Is your site in the 12%? Our web design work starts from exactly this: the call button, the quote form and the path between them. If you'd rather keep your current site, we can add both to it.
Finding 4: only 7% pass Google's mobile speed test
We ran every homepage that loaded through Google's PageSpeed Insights, mobile setting, on Google's own servers.
| Mobile PageSpeed result (479 homepages) | Share |
|---|---|
| Good (90 to 100) | 7% |
| Needs improvement (50 to 89) | 66% |
| Poor (0 to 49) | 27% |
| Median score | 61 |
| Median Largest Contentful Paint (lab) | 8 s |
| LCP slower than 4 seconds | 79% |
| Median page weight | 2.6 MB |
| Heavier than 3 MB | 44% |
The lab test is a simulation. The real-user numbers are more forgiving, but only a minority of these sites have them: Google's Chrome UX Report had enough real visitors to report field data for 24% of the sites. Of those, 62% load their main content within Google's 2.5-second "good" line for real phone users, and 52% pass all three Core Web Vitals.
The weight isn't mostly photos, which surprised us. On the median homepage, images came to 594 KB while JavaScript came to 885 KB, and 54% of all bytes were loaded from other companies' domains: builder CDNs, font services, chat and review widgets, booking tools and trackers. Every one of those is a request the phone has to make before the page settles. If your site is in the red, our guide to improve website speed lists the fixes in the order we do them.
Speed by platform
| Platform | Sites tested | Share of live sites | Median mobile score | Median page weight |
|---|---|---|---|---|
| WordPress | 205 | 42% | 60 | 2.7 MB |
| Custom or unidentified | 98 | 21% | 65 | 2 MB |
| Wix | 38 | 8% | 67 | 2.9 MB |
| GoDaddy Builder | 30 | 6% | 65 | 1.8 MB |
| Squarespace | 30 | 6% | 49 | 2.6 MB |
| Shopify | 17 | 4% | 60 | 3.4 MB |
| Weebly | 12 | 3% | 45 | 2.1 MB |
| Other | 11 | 2% | 61 | 6.3 MB |
| Next.js | 10 | 2% | 72 | 1.9 MB |
| Duda | 9 | 2% | 55 | 5.2 MB |
| Webflow | 10 | 2% | 56 | 8.9 MB |
| Vertical builder | 8 | 2% | 58 | 1.6 MB |
We don't read this table as "platform X is fast". A well-built WordPress site can score in the 90s and a careless custom build can score 20. Several platforms have only 8 to 12 sites here, so treat those rows as a hint, not a ranking. What the table does show is where the typical small business site lands when nobody is watching the numbers. When we build on Next.js, speed is part of the build, not an afterthought.
Finding 5: 59% don't tell Google what kind of business they are
LocalBusiness structured data is a short block of code that says "this is a dentist at this address, open these hours, with this phone number". Google documents it for exactly this kind of business.
| Structured data on the homepage | Share of live sites |
|---|---|
| Any schema at all | 71% |
| LocalBusiness or a specific type (Dentist, LegalService, AutoRepair...) | 42% |
| Review or rating markup | 12% |
| FAQ markup | 8% |
The gap between those two rows is mostly generic markup: of the 147 homepages with schema but no local type, 105 carry only WebSite, WebPage, Organization or breadcrumb blocks, the kind SEO plugins and builders add automatically. The part that says what the business actually is, where it is and when it's open is missing on 59% of homepages. That's a gap in local SEO basics, and one of the easiest to close.
Finding 6: 28% of footers say 2024 or earlier
A stale copyright year doesn't hurt rankings on its own. It's a signal to the visitor that nobody has touched the site in a while, and it usually is true. Of the 303 homepages that show a year, 28% say 2024 or earlier and 18% say 2022 or earlier. The oldest footer in the sample still says 2008.
Other signs of neglect we measured on live homepages:
- 25% have no meta description, so Google writes the search snippet for them.
- 22% have no H1 heading, and 20% have more than one.
- 74% have at least one image with no alt text.
- 39% have no Google Analytics or Tag Manager code, so the owner can't see how many visitors the site gets.
- 13% load an accessibility overlay widget, which does not make a site accessible by itself.
Finding 7: owners rarely block AI crawlers. Firewalls sometimes do.
This is the finding we'd most like other people to test. The common advice for "getting found in ChatGPT" starts with checking robots.txt for blocked AI bots. On local business sites, that's almost never the problem.
| AI crawler access, 489 live homepages | Share |
|---|---|
| robots.txt names any AI crawler | 11% |
| robots.txt blocks any AI crawler | 4 sites (0.8%) |
| Refused a request with the ClaudeBot user agent | 13% |
| Refused a request with the GPTBot user agent | 7% |
| Refused a request with the PerplexityBot user agent | 4% |
| Refused at least one of the three | 14% |
Every one of those sites answered a normal browser request fine a moment earlier. The refusals come from the firewall, CDN or hosting layer, not from a line the owner wrote in robots.txt. 54% of them sat behind Cloudflare.
An honest caveat: our requests came from our own servers, not from OpenAI's or Anthropic's published IP ranges. Some firewalls let the real crawler through and only block impostors, which is sensible. So a refusal in our test is a warning sign, not proof. If you care about AI search visibility, check your firewall's bot settings, not just robots.txt. OpenAI documents its crawlers and IP ranges on its bots page.
llms.txt: 31% have one, and most owners didn't write it
151 of the 489 live sites serve an llms.txt file. That sounds like early adoption until you split it by platform: 31 of 38 Wix sites and 16 of 17 Shopify sites have one, against 49 of 205 WordPress sites.
We downloaded the files again and looked at who made them: 73% carry the fingerprint of a website builder (Wix, Shopify, Duda, GoDaddy) or an SEO plugin (Yoast, AIOSEO, Rank Math). The other 28% carry no fingerprint either way; some of those are generated too. Google has said its search systems don't use llms.txt, and no AI assistant has confirmed it relies on the file, so we'd treat it as optional. The things that matter more for answer engine optimization are the basics in findings 1 to 5.
Which industries do best and worst
| Industry group | Listed | Live | No tap-to-call | No form or booking | LocalBusiness schema | Median mobile score |
|---|---|---|---|---|---|---|
| Health & dental | 201 | 156 | 14% | 15% | 44% | 59 |
| Beauty, fitness & personal | 150 | 107 | 36% | 26% | 34% | 62 |
| Legal, finance & real estate | 120 | 91 | 25% | 15% | 55% | 59 |
| Trades & home services | 105 | 81 | 27% | 40% | 38% | 64 |
| Auto | 73 | 47 | 38% | 47% | 36% | 59 |
Health & dental businesses do best on contact options: only 14% lack a call link and 15% lack a form or booking. Auto businesses do worst on both: 38% have no tap-to-call link and 47% have no form or booking.
The 10-minute check for your own local business website
Every item below comes from a finding above. If you can tick all ten, you're ahead of most of the 662 sites we tested.
- Click the website link on your Google Business Profile, Yelp and Apple Maps listings. Does it load, on HTTPS, with no warning?
- Open your homepage on your phone. Tap the phone number. Does it dial?
- Is there a form or booking button above the fold, or one tap away?
- Send yourself a test request. Did it arrive, and in whose inbox?
- Run your homepage through PageSpeed Insights on mobile. Is it 50 or higher? Is the main image under 200 KB?
- Paste your homepage into Google's Rich Results Test. Does it find LocalBusiness (or your specific type)?
- Does the footer show the current year?
- Does the homepage have one H1 that says what you do and where?
- Is Google Analytics or Tag Manager installed, and does someone look at it?
- Ask your host or developer whether the firewall blocks AI crawlers, and whether that's on purpose.
Rather have someone else run the list? We'll check all ten on your site and send you the results with fixes in order of impact. Get your free website and SEO audit.
What RankMeHi does about this
We're a web design and development team in Costa Mesa, and most of what we fix for local businesses is on the list above. Three ways we usually help:
| If your site... | What we do | Starts at |
|---|---|---|
| Works but leaks calls and requests | Audit and fix: call links, forms, schema, speed, tracking | $490 |
| Is slow, dated or on a builder you've outgrown | New fast website, 5 pages, design included | $2,490 |
| Is fine and you want more from it | Ongoing SEO and conversion work | $690/month |
| Needs a new site but not a big upfront bill | Pay-per-lead: we build it, you pay per approved lead | $1,490 setup, then from $35 per lead |
When you don't need us: if your site loads fast, has a working call link and form, and you're getting the calls you can handle, leave it alone. A redesign won't fix a problem you don't have.
Key takeaways
- About 1 in 7 websites linked from local business listings in Orange County doesn't work at all.
- 26% of live homepages have no tap-to-call link and 25% have no form or booking.
- Only 7% score 90+ on Google's mobile PageSpeed test; the median is 61.
- 59% are missing LocalBusiness schema.
- AI crawler blocks come from firewalls far more often than from robots.txt, and llms.txt files are mostly platform-generated.
Methodology and limitations
Sample. All businesses in Orange County, California (OpenStreetMap relation for the county, via the Overpass API) with a `website` tag and a local service category: crafts and trades, repair shops, beauty and personal care, offices (legal, finance, real estate, insurance, contractors), health (dentists, doctors, clinics, veterinarians, chiropractors, optometrists) and fitness. Chains were removed: anything with a brand tag, and any domain shared by four or more locations. Social media, Yelp and Google links were removed.
Result: 662 unique domains. OpenStreetMap is volunteer-maintained, so the sample leans toward businesses that someone has mapped, and some listings are out of date. A few chain locations without brand tags may remain.
When. Crawled 2026-09-25 to 2026-09-26. PageSpeed Insights runs 2026-09-26.
Reachability. Each homepage was requested with a current Chrome user agent, following redirects, with a 20-second timeout. DNS was resolved through public resolvers, and every "domain not found" was confirmed with Cloudflare DNS-over-HTTPS. Firewalls that showed a challenge page or refused a normal browser request (401, 403, 405, 429) were counted as "not measured", not as broken.
Page checks. Done on the homepage HTML. Sites where the raw HTML was missing a call link, form or schema were loaded again in headless Chrome with a phone viewport and re-checked after JavaScript ran.
That re-check queued 359 sites and 255 rendered successfully; the rest keep their raw-HTML result, so the "missing" numbers may be slightly high. A call link means an `href="tel:"` link. A form means a `<form>` element or a known form or booking embed. Schema types were read from JSON-LD.
Speed. Google PageSpeed Insights API, mobile strategy, Lighthouse 13.5.0, one run per homepage, plus Chrome UX Report origin-level phone data where available. Single lab runs vary by a few points; the distribution is more reliable than any one score.
AI crawlers. robots.txt was parsed for named AI user agents (GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, anthropic-ai, PerplexityBot, Google-Extended, CCBot, Applebot-Extended, Bytespider, meta-externalagent). The homepage was then requested once each with the GPTBot, ClaudeBot and PerplexityBot user agents from our own IP. A refusal counts only as an explicit 4xx response or a challenge page; rate limits (429), server errors and timeouts were not counted.
Privacy. We don't name any business, and we won't. The aggregate numbers are free to use.
Use these numbers
You're welcome to quote any number or embed any chart from this study in your own article, deck or newsletter. Please credit "RankMeHi Local Business Website Study (2026)" with a link to this page. If you want a cut we didn't publish (one city, one industry, one platform), email us and we'll run it.
FAQ
What percentage of local business websites don't work?
In our September 2026 test of 662 Orange County business websites listed on public map data, 15% did not work: the domain was gone, the server returned an error, the SSL certificate was broken, or the domain was parked. Another 11% were behind firewalls we could not measure.
How many small business websites are mobile friendly?
94% of the live homepages we tested have a mobile viewport tag, so nearly all of them at least resize for phones. Being usable on a phone is different: 26% have no tap-to-call link and only 7% score "good" on Google's mobile PageSpeed test.
Do small business websites block ChatGPT and other AI crawlers?
Very few do it on purpose. Only 4 of the 489 live sites in our study (0.8%) block any AI crawler in robots.txt. But 14% refused at least one request carrying an AI crawler user agent, usually at the firewall or CDN level. Some of those firewalls may still allow the verified crawlers.
Does my local business website need an llms.txt file?
Not yet. 31% of the sites we tested have one, mostly because Wix, Shopify and some WordPress plugins create it automatically. No major AI company has said it relies on the file. A working site, clear contact options and LocalBusiness schema matter more today.
What is the most common problem on local business websites?
In our data, missing LocalBusiness schema (59%) and images without alt text (74%) are the most widespread. The costliest are the ones that lose leads directly: a dead website link (15% of listings), no tap-to-call link (26%) and no form or booking (25%).
The short version
Most local business websites in Orange County aren't failing because of design trends or AI. They fail at the basics: the link is dead, the phone number doesn't dial, there's no way to ask for a quote, and the page is slow on a phone. Those are fixable in days, not months.
Find out where your site stands. We'll run every check from this study on your website and send you the results in plain English, free. Book your free website and SEO audit.
Services mentioned in this article
RankMeHi designs and builds business websites, then makes sure they are found and convert. Start with the two pages below.