Microscope examining a single small building block on its stage, a picture of AI visibility audit findings

AI Visibility Audit: What 41 Local Business Sites Revealed

An AI visibility audit checks whether AI tools like ChatGPT, Gemini and Google's AI Overviews can read your website, believe what it says, and recommend you to a customer. In August and September 2026 we ran one on 41 local businesses across North America. Every one of them explained its services clearly. Not one scored above 41 out of 100, and every site had urgent fixes waiting. What held them back was small, visible and fixable.

Key Takeaways

  • The average score across our 41 audits was 34 out of 100, and the best score anyone reached was 41.
  • Access was mostly fine: 71% of the sites passed the crawl-access check. Trust is where they lost their points.
  • Two out of three of the businesses we audited wrote their own business name more than one way across their website and public profiles.
  • Five out of six of them had an error a customer could see on the page, from typos to leftover template text.

If you have asked ChatGPT for a plumber or a roofer in your own town and your company was missing from the answer, this page shows what our AI visibility audit found on 41 real local service websites, with a check you can run on your own site under each finding. I've worked in search for more than fifteen years, and these audits still surprised our team: most sites let AI in the door, then lost its trust over details like a misspelled service or a second phone number.

What is an AI visibility audit?

An AI visibility audit is a review of the signals AI tools use to decide whether your business is real, what it does, and whether it is safe to recommend. Ours looks at six areas:

In our AI gap report analysis of 17 local businesses across 1,352 AI answers, we found AI named the business being asked about in just 18.1% of answers, 245 out of 1,352. 2 of the 17 businesses were never named at all, by any engine, in any city.

  • Whether AI crawlers can reach your pages at all, through your robots.txt file (a short file that tells bots where they may go) and your sitemap (a list of your pages).
  • Schema, which is labeling in your page code that tells machines what each page is about.
  • Whether your name, phone number and other facts match everywhere they appear.
  • Proof a machine can verify, like reviews, named people and finished work.
  • How your content is organized.
  • Anything visible on the page that undercuts trust.

When I open a new audit, I check two things first. Can AI crawlers get in? I read robots.txt for blocks on AI bots, confirm that an llms.txt file (a plain-text guide to your site written for AI tools) actually loads, and make sure the sitemap lists real pages. Then, does the business write its name the same way on its site and its profiles? On these 41 sites, crawler access mostly passed and the name check mostly failed.

Each audit ends in a written AI visibility report and an AI visibility score from 0 to 100. That score measures something different from the free online checkers. Those tools count how often an AI mentions a brand. Ours measures whether your website gives AI a reason to mention you in the first place.

This matters more every month. BrightLocal's Local Consumer Review Survey 2026 found that the share of consumers using AI to find local businesses climbed from 6% in 2025 to 45%. If the idea is new to you, our guide to AI visibility covers the basics. The 41 businesses in this study were mostly plumbers, spread across at least 18 US states and one Canadian province.

If you want your own numbers, our free audit runs these same checks.

What is a good AI visibility score?

No business in this study scored above 41 out of 100, so on our scale anything past that mark beats every site we audited. A score in the mid-30s puts you in the middle of the group, and the useful part of any score is the ranked list of fixes that comes with it.

Every one of the 41 audits rated service clarity as strong, and the average score was still 34.2 out of 100. The median was 35. The lowest score was 24 and the highest was 41. These owners knew what they sold and said so plainly. A clear description of services was the one thing every site already had, which makes it the one thing none of these audits needed to fix. The trouble sat somewhere else.

The scores bunched low. Of the 41, 28 (68%) scored 35 or below, and 38 (93%) scored under 40. The 28 plumbing companies in the group averaged 34.9, a touch higher and still well short of halfway.

Each of these businesses came to us as a prospect in August or September 2026, and we ran the same comprehensive AI Visibility Audit on every one. With one audit across all 41, and 28 of them in one trade, the same gap showed up across the whole group: clear services, weak trust. Even the top scorer earned no positive rating on AI trust signals.

Every site also had urgent fixes, the problems we flag to correct before they cost a business trust. Across the 41 audits we logged 208 of them, an average of about five per site, and every site had at least three.

If a score in the 30s sounds like the setup for a sales pitch, what actually arrives is a written report and a ranked list of urgent fixes, most of which you can make yourself in an afternoon.

Here is the part that surprised us. Getting AI to the website was mostly handled. Of the 41 sites, 29 (71%) passed our crawl-access check, and 33 (80%) pointed bots to their sitemap correctly. Our piece on crawler access explains the two files involved if you want the mechanics.

Trust told a different story. Only 5 sites (12%) earned a strong rating for technical AI readability overall. Not one of the 41 earned a positive rating on AI trust signals: 38 were rated "needs work" and 3 were rated "bad." The door was open. What AI found inside gave it little reason to vouch for anyone.

That pattern repeats through every finding below. Access is mostly solved for these businesses. Belief is where they lose, and the sections that follow show exactly where it leaks.

The fastest way to see where your own site lands is a short call to look at it together.

How many businesses wrote their own name more than one way?

Of the 41 businesses, 28 (68%) wrote their business name more than one way across their website or public profiles. On their own websites alone, 22 (54%) did it. Among the plumbers, 17 of 28 (61%) had their name written two or more ways on their own site.

It rarely looked dramatic. An ampersand in the header and the word "and" in the footer. "Inc." on the About page and nothing on the homepage. A short nickname in the body copy. A parent company's name on one page and the trade name on the next.

It is easy to treat these as cosmetic, because every customer knows the ampersand version and the "and" version are the same company. A machine reading your site cold cannot be sure. Entity hardening is how we remove the doubt: drawing clear boundaries around what your business is, what it does and which services it does not offer, so AI tools form one accurate picture of you.

Why does this matter to AI? An AI system keeps one record for each business, called an entity: a single thing with a name, a location and a set of services attached. When your site calls you three different things, it hands that system three candidates and no way to be sure which one is you. Our guides to name consistency and entity SEO go deeper on how that record gets built.

The same split happens across websites. Six of the businesses (15%) spread one brand over more than one domain, which divides the same signals between two addresses.

This check takes five minutes. Search your own site for every version of your name, then pick one. To keep it from drifting again, set up a single source of truth: one document holding your exact business details, which every page, profile and listing copies from.

Pick one exact name and use it everywhere, including the code layer machines read, and check the rendered page code, not only the words a visitor sees.

A quick call is all it takes to see how your own site reads to AI.

Are typos bad?

For the buyer reading your page, yes. A misspelled service or a line of leftover template text tells a stranger nobody is minding the site, and on these 41 sites that kind of error was common.

Thirty-four of the 41 sites (83%) had at least one visible copy error: a typo, placeholder text, stray characters or garbled symbols that any customer could see.

Spelling and grammar were the most common. In their own copy, 17 of the 41 (41%) had spelling or grammar errors. Counting errors inside the testimonials they chose to display, the number rises to 23 (56%).

Leftover template text was the most striking. Nine sites (22%) still had placeholder text live on the public site. A Florida plumber had a service page showing lorem ipsum, the fake Latin filler designers use in mockups, with "Author Title" printed where customer testimonials belong. A Kentucky plumber's page still carried the website builder's default line: "This is a paragraph where you can add any information you want to share with website visitors." A Florida restoration company had the words "New Paragraph" sitting on its homepage.

We made this mistake too: our own site had a designer's internal note live under a heading on about 23 pages until our September 2026 audit caught it, and every AI engine could read it. So run a machine-style crawl of your live site before you treat any copy as finished.

Some errors landed on the very words the business sells. A Missouri plumber's footer misspelled one of its own services and repeated the misspelling three times.

We did not measure whether errors like these change what AI recommends, so we will not claim they do. What we can say is simpler. The customer reads the same page the AI reads. A buyer judging website credibility sees placeholder text and a misspelled service and draws the obvious conclusion: nobody is looking after this site. That impression lands before the phone ever rings.

The check: read your footer and your full service list out loud, slowly. Typos hide in the places you stopped reading years ago.

If you would like a second set of eyes on your pages, get your audit.

What happens when a website contradicts itself?

When a website contradicts itself, neither the customer nor the AI can tell which fact is real, and in one case below it sent callers to a number the page never showed.

Eleven of the 41 businesses (27%) published facts that contradicted each other, either on their own site or between their site and their Google listing. The conflicts covered review counts, phone numbers, license numbers, addresses and email addresses.

Review counts were the most common clash. Seven sites (17%) showed conflicting review numbers, and all seven were plumbers, which works out to 25% of the plumbing group.

Each example sounds small on its own. A Florida plumber showed two different review counts on the same site and listed three different license numbers. A California plumbing company's homepage claimed a few hundred Google reviews in one spot and nearly two thousand in another. An Oklahoma plumber called itself BBB (the Better Business Bureau) accredited while its BBB listing said otherwise.

One conflict costs calls directly. A Colorado company displayed one phone number, but its "Call Today!" button dialed a different one. Every customer who tapped that button reached a number the page never showed them.

A system that finds two answers to the same question cannot tell which one is true, so it has good reason to trust neither. People react the same way. Two review counts on one page read like a number nobody checked.

The check: open your site on your phone and tap every phone link. Then compare every review count and license number on the site with each other and with your Google listing.

When you are ready to see what your own site shows, book a call.

What are trust signals on a website?

Trust signals are the proof a stranger, or an AI tool, can check for themselves: distinct testimonials, finished projects, named people, a guarantee that says what it covers and a clear path to your reviews.

Proof is what separates a business that says it does good work from one that shows it. On these sites, much of that proof looked recycled or was missing entirely.

Start with testimonials. Of the 41, 13 (32%) repeated the same testimonials on one page or across several pages, and among plumbers the rate was 36%. A buyer who meets the same three quotes on the homepage, the service page and the contact page stops believing any of them. Fresh, specific real reviews carry more weight than a polished trio on repeat.

Finished work was missing almost everywhere. Of the 41, 35 (85%) had no case-study, results or project page at all. For a picture of what that kind of proof can look like once it is written down, see our local SEO case study.

People were missing too: 27 businesses (66%) showed no leadership or team bios. Person schema, the code label that tells machines who is behind a business, was present on only 3 sites (7%). An AI system deciding who to trust had no one to point to.

Claims ran ahead of evidence. Nineteen sites (46%) stated a guarantee or warranty, and only 7 (17%) said what it actually covers. Ten (24%) made unqualified claims such as "#1," "the best" or "100% guaranteed" with nothing on the page backing them, and 5 of those (12%) called themselves #1 or the best outright.

Outside proof was hard to confirm. For 26 of the 41 (63%), the audit could not verify their review profiles from the website itself. Those reviews may well exist. The site simply gave neither a machine nor a buyer a clear path to find them.

We run this audit free, and it checks every item above for your own site.

How does AI choose its sources?

For the questions buyers ask before they call, an AI assistant leans on the page that already answers the question in plain words, so the businesses it can quote are the ones that gave those questions a home.

Buyers ask the same handful of questions before they pick up the phone, and most of these sites gave those questions no home.

That gap matters because of how an assistant answers. When a customer asks it something, it looks for a page that already answers the question in plain words. A clear question as a heading, with a two-sentence answer under it, hands the assistant a ready quote. A site without one leaves it to quote somebody else.

FAQ pages were rare. Only 8 of the 41 (20%) had a dedicated FAQ page, and 22 (54%) had no FAQ content anywhere. Of the 19 that answered questions somewhere, only 4 marked those answers with FAQ schema so machines could read them as questions and answers. Clear question-and-answer content is one of the most direct ways to get cited by an assistant.

Price was a blank. Of the 41, 36 (88%) published no pricing at all. When a customer asks an assistant what a job costs, these sites gave it nothing of their own to quote.

Leaving prices off can look like the safe choice, because no two jobs are quite the same and a number on the page can read like a promise. But when a buyer asks an assistant what the job will run, a site with no price of its own leaves the assistant to quote somebody else's number.

The labeling was thin. Organization schema, the label that tells machines who a business is, was present on 20 of the 41 (49%) and partial on 3 more. For 12 sites (29%), the audit confirmed none of the six core schema types it checks. Our plain-English guide to schema markup shows what those labels look like.

Book a call and we will walk through which questions your site already answers.

How to improve AI visibility of a website

Start by fixing the facts AI already has reason to doubt, before you add anything new.

The good news in this data is that the most common problems are also the cheapest to fix. In order:

  1. Create a single source of truth for your business details, and have every page, profile and listing copy from it.
  2. Proofread your footer, service list and homepage, and delete any template text.
  3. Make every phone number, review count and license number match, on the site and on Google.
  4. Swap repeated testimonials for distinct ones, and put a named person behind the business.
  5. Answer your top buyer questions on the page, then add schema so machines can read them.

Treat that first item as a style guide for your business identity: your exact name, address, phone, hours, services and logo in one place, copied from there every time.

Most of that list is an afternoon of work, and all of it belongs ahead of any new content spend. Once it is done, you can check whether AI recommends you and see what changed.

New content is the tempting first move, because a new article looks like progress. But it sits on the same site as any second spelling of your name or phone link that dials the wrong number. I would finish these five fixes before publishing a single new article, and only start on content once the checks come back clean.

When we do this work for a business, it runs in three steps. Our comprehensive AI Visibility Audit reads data points from inside your site's code, such as schema and whether pages answer questions up front, to show where assistants misread or skip you. Then we make the core facts consistent everywhere and structure the data so crawlers find clean answers. Then we track incoming calls and booked jobs and keep adjusting until those numbers move.

Get your audit to see this list ranked for your own site.

How we ran these audits, and what the scan cannot see

These findings come from 41 AI Visibility Audits that GetLocalLeads.AI, an AI visibility and digital marketing agency for local and multi-location service brands, prepared in August and September 2026: 39 in August and 2 in September. Each one reviewed a business's public website. These were businesses we audited, not clients we served, so nothing here describes results we produced.

Twenty-eight of the 41 were plumbing companies, some also offering HVAC. The rest covered pest control, roofing, landscaping, water treatment, restoration, appliance retail and several professional and personal services. Forty were in the United States and one was in Canada. Every percentage in this piece is a share of those 41 audits, not a claim about all local businesses.

The scan has limits, and we would rather name them. It reads public pages, so it cannot always confirm a Google Business Profile or a third-party review count. A few checks were not reported for every site, so we left them out. Duplicate navigation text that may come from the scanning process itself was excluded from every count.

Frequently asked questions

What does an AI visibility audit check?

An AI visibility audit checks whether AI tools can reach your website, understand what it is about and trust what it says. Ours covers crawler access, schema, name and fact consistency, verifiable proof, content organization and visible trust problems. It ends with a score out of 100 and a ranked list of fixes.

Do I need an AI visibility audit if I already rank on Google?

Ranking and being recommended are separate questions, and our audits measure the second one: whether your site gives AI a reason to trust it. That question matters more each year. BrightLocal's 2026 survey found 45% of consumers now use AI to find local businesses, up from 6% a year earlier.

What is a good AI visibility score?

In our 41 audits, no business scored above 41 out of 100, and the average was 34. A low score is normal and useful, because it arrives with a ranked list of what to fix. Free online checkers that show an AI visibility score usually measure something different: how often a brand gets mentioned.

How much does an AI visibility audit cost?

GetLocalLeads.AI runs a Free AI Visibility Audit. To qualify, you need a website, a name, an email and a company URL. The comprehensive audit carries a stated value of $997, and you can request yours by booking a call.

Can I check my own website for these problems?

Yes. The five fixes above double as a self-check you can run in an afternoon. To test whether assistants actually name you, follow our guide to checking whether AI recommends your business.

The pattern across these 41 businesses is oddly hopeful. The problems that cost them the most trust were the ones that cost the least to fix: a name, a typo, a phone link, a repeated quote. The fix starts with someone reading the site the way a stranger would, and our AI visibility services pick up from there. If you want that first read done for you, book a call and we will run your free audit.

More GetLocalLeads.AI case studies