We X-rayed 240 Phoenix small business websites. Here is what Google and AI can't see.

Most small business owners only see their website the way a customer does, in a browser, on a good connection. Search engines and AI assistants see something different. They read the code, often without running the scripts that fill in the page, and they look for specific signals: what the business does, where it is, how to reach it.

We wanted to know how Phoenix-area businesses hold up. So we ran 240 of their websites through Site X-ray, the free tool we built that loads a page on a simulated phone and reads it the way a crawler does.

The short version

• 38% of sites had at least one gap that makes them harder for search engines and AI assistants to read.
• 31% had no structured data at all, the code that tells Google your business name, type, address and hours.
• 35% of local businesses had no tap-to-call link, so phone visitors have to copy the number by hand.
• 33% showed no call, book or quote button on the first phone screen.
• 24% loaded their main content within Google's 2.5 second target on a simulated 4G phone.
• 1% blocked AI crawlers. The worry that small businesses are turning AI away doesn't match what we found.

How we picked the sites

We started from public OpenStreetMap listings in the Phoenix metro that include a website: restaurants, cafes, dentists and clinics, salons, gyms, shops, contractors and professional offices from Phoenix, Mesa, Scottsdale, Tempe, Gilbert, Chandler, Glendale, Peoria, Surprise and nearby towns.

We removed national chains, hospitals and government offices, then picked 300 businesses at random. Of those, 240 loaded and were scanned on October 13, 2026. Ten sites blocked automated visitors, a handful returned errors, and at least 31 of the 300 websites (about 1 in 10) didn't load at all when we checked them in a normal browser. Those are businesses still listed online with a website that no longer works.

Each site got one automated visit on a simulated 4G phone. We publish only totals, never individual business names.

What search engines and AI assistants miss

AI assistants like ChatGPT, Claude and Perplexity increasingly answer questions such as "who's a good plumber in Mesa?" They can only recommend a business they can read and understand. Here is where Phoenix sites fell short.

• No structured data: 31% of sites. Structured data is a small block of code that states your business name, type, address, phone and hours in a format machines read directly. Among sites that look like local businesses, 58% lacked the local business version of it.
• Thin or hidden text: 19% of sites. These sites gave crawlers fewer than 150 words, or hid at least half their words behind scripts that some crawlers don't run. One in ten gave crawlers fewer than 50 words.
• Missing basics in search results. 17% had no meta description, so Google picks random text for the search snippet. 27% had no main headline in the code. A third had a page title that was missing, too short or too long.
• No sitemap: 15% of sites. A sitemap tells search engines which pages exist.

Two findings ran against what we expected. Only 3 of 240 sites blocked any AI crawler, and none blocked search engines entirely. And 23% already had an llms.txt file, a newer format meant to help AI tools. Some of the ones we opened appear to have been generated automatically by the website platform.

What phone visitors miss

Many local searches happen on a phone, and those visitors usually want to call, book or get directions.

• No tap-to-call link: 35% of local businesses. For restaurants, cafes and bars it was 55%.
• No call, book or quote button on the first screen: 33%. The visitor has to scroll to find a way to act.
• Hours or address not in the page text: 37% of local businesses. Crawlers can't read details that appear only inside an image or a map embed.
• No contact form, booking or online store: 32%. After hours, the only option is a phone call that nobody answers.
• No reviews on the homepage: 46%.
• No chat or site assistant: 89%.

Speed and weight

On a simulated 4G phone, the median site showed its main content in 3.5 seconds. Only 24% hit Google's 2.5 second target. Half the pages weighed more than 2 MB, and 9% weighed more than 5 MB. Heavy pages take longer to show up on a phone connection.

Speed numbers come from one automated visit and move from run to run, so treat them as a direction, not a precise measurement.

How each industry scored

Site X-ray scores each site out of 100. The median across all 240 sites was 72.

• Trades and home services (21 sites): median score 83. Only 10% of local contractors lacked tap-to-call.
• Offices and professional services (44 sites): median 79.
• Health, dental and wellness (37 sites): median 77. None had a thin-text problem.
• Salons, fitness and personal care (16 sites): median 69. Half had no structured data.
• Shops and retail (36 sites): median 67. 44% had no structured data, and half of local shops had no tap-to-call link.
• Restaurants, cafes and bars (86 sites): median 65.5, the lowest. 55% of local spots had no tap-to-call link, and 30% gave crawlers thin or hidden text.

What to fix first

Most of these are small jobs. In rough order of effort:

1. Make your phone number a tap-to-call link. It's one line of code, or a setting in most website builders.
2. Put your hours and address in plain text on the homepage or contact page, matching your Google Business Profile.
3. Write a real meta description and page title that name your service and your city.
4. Add one clear button at the top: Call, Book or Get a quote.
5. Add local business structured data with your name, address, phone and hours. Many platforms have a setting or plugin for it.
6. Shrink your photos. Large images are a common reason pages get heavy.
7. If crawlers see fewer than 150 words, add a few short paragraphs about what you do, where and for whom, as real text rather than images.

Check your own site

Run the free Site X-ray to see your score and fix list in about 30 seconds, with no sign-up. If you'd rather have someone walk through it with you, book a time with AJ.

Method notes: sample drawn from OpenStreetMap listings with a website tag in the Phoenix metro on October 13, 2026, excluding national chains, hospitals, government and religious organizations. 300 sites sampled, 240 scanned. Scans ran with Site X-ray's checks from a single test server with one visit per site on a simulated 4G phone. Business category comes from the OpenStreetMap listing. Percentages for local-business checks use only sites Site X-ray classified as local businesses (201 of 240).