I ran a test last month that I’ve been thinking about ever since.

I asked ChatGPT, Perplexity, Google AI Mode, and Claude to recommend businesses across ten categories in Tulsa. Plumbers. Electricians. Restaurants. Dentists. Real estate agents. Auto repair shops. Accountants. Personal injury attorneys. HVAC companies. And digital marketing agencies.

Forty queries total. Ten categories, four AI platforms each.

Across all forty queries, the AI systems recommended a total of about five to seven businesses per category. The same names kept showing up. The same plumber across three platforms. The same dentist across two. The same restaurant across all four.

For each category, there are dozens of businesses in Tulsa that offer the same service. In some categories, hundreds. The AI systems recommended less than five percent of them. The rest didn’t exist. Not ranked lower. Not mentioned with a caveat. Completely absent from the answer.

This lines up with the national data. Only 1.2% of local service businesses appear in AI assistant recommendations compared to 35.9% in Google’s traditional local pack results. AI systems are 30 times more selective than Google search. And the gap between being recommended and being invisible isn’t a ranking position. It’s a binary. You’re either in the answer or you’re not.

Most local business owners think they’re doing fine because they show up in Google Maps and the local pack. They have no idea that 45% of consumers are now using AI tools to find local services, up from 6% one year ago. And in that AI channel, their business doesn’t exist.

I. Why AI Systems Are So Much More Selective

Traditional Google search shows ten results per page. The local pack shows three, sometimes four. Even if you’re not in the top three, you’re on the page somewhere. A motivated searcher can scroll, click “More places,” and find you.

AI search doesn’t work that way.

When someone asks ChatGPT “who’s the best plumber in Tulsa,” it doesn’t show ten results. It gives a recommendation. Maybe two or three names with brief explanations of why each one is a good choice. That’s it. There’s no page two. There’s no “More places” button. There’s no scrolling to find alternatives. The AI made a decision and presented it as an answer.

The selection criteria are fundamentally different from traditional search. Google’s local algorithm weighs proximity, relevance, and prominence. The prominence factor includes review count, review score, and backlinks. It’s a relatively formulaic system that rewards businesses for checking specific boxes.

AI systems evaluate entity authority. They’re not just checking whether your Google Business Profile is filled out. They’re synthesizing information from across the entire web to determine whether your business is a credible, recognized entity worth recommending. They cross-reference your website, your reviews, your social media presence, your mentions in other contexts, your structured data, and the consistency of your information across platforms.

A business with 200 Google reviews but no website updates in two years, no social media activity, inconsistent NAP data across directories, and no schema markup might rank fine in Google’s local pack. That same business is invisible to AI systems because the entity signals are weak and inconsistent. The AI can’t confidently recommend a business it can’t verify is current, active, and legitimate.

This is the fundamental shift that local businesses are missing. Google’s local pack rewards optimization. AI search rewards authority. They look similar on the surface but they’re measuring completely different things.

II. The 4.0 Star Threshold Nobody Told You About

Here’s a data point that stopped me when I first saw it.

AI assistants typically exclude businesses below 4.0 stars regardless of other factors. The businesses that get recommended average 4.1 to 4.3 stars with active review responses.

Let that sink in. Below 4.0 stars and you’re essentially invisible to AI recommendations. Not ranked lower. Excluded. The AI doesn’t recommend businesses with mediocre reviews because recommending a 3.5-star plumber to someone asking for “the best plumber” would make the AI look bad. These systems optimize for user satisfaction with their recommendations. A bad recommendation erodes trust in the AI platform itself.

And it’s not just the star rating. AI systems look at review velocity and recency. A business with 500 reviews but nothing in the last six months sends a different signal than a business with 200 reviews and 15 in the last month. The freshness signal tells the AI whether the business is currently delivering good experiences or whether those reviews reflect a past version of the business.

Review responses matter too. When a business owner responds to reviews, positive and negative, it signals active management. It shows the AI that the business is engaged and responsive. Businesses that respond to reviews appear in AI recommendations at noticeably higher rates than businesses that don’t, even when the star ratings are similar.

Most local business owners think of reviews as a reputation management exercise. In 2026, reviews are an AI visibility input. They’re one of the primary signals AI systems use to determine whether to recommend your business to someone who just asked for help.

III. Your Google Business Profile Is an AI Input Now

The Google Business Profile used to be a listing. Fill out the fields, add some photos, maybe post an update once in a while. It showed up in Maps and the local pack. Simple.

In 2026, your GBP is one of the primary data sources AI systems use to understand your business. Google AI Mode pulls directly from GBP data. ChatGPT and Perplexity cross-reference GBP information with your website and other platforms. The information in your GBP isn’t just showing up in Maps anymore. It’s feeding the AI systems that decide whether to recommend you.

Category selection matters more than ever. AI systems use your primary and secondary categories to classify your business entity. Your categories need to be precise and comprehensive.

Your business description needs to be written for machines, not just humans. AI systems parse your description to extract entity information. The description that says “We’ve been proudly serving the Tulsa community since 1987” tells the AI nothing. The description that says “Data-first digital marketing agency specializing in AI search visibility, entity authority strategy, and marketing analytics for mid-market businesses in Oklahoma” gives the AI structured entity information it can work with.

Photos and posts signal freshness and activity. Businesses that post new photos or updates at least twice weekly show significantly higher recommendation rates. Not because AI systems look at your photos. Because posting frequency is a proxy for business activity. A business that hasn’t posted to GBP in three months might be closed. The AI doesn’t know. So it recommends the business that posted yesterday instead.

Products and services sections need to be comprehensive. AI systems use these structured data fields to match businesses with specific queries. If someone asks “who does website design for small businesses in Tulsa” and your GBP lists “website design” as a service with a description, you’re matchable. If you don’t have it listed, you’re invisible for that query even if it’s your core offering.

Q&A sections are entity signals. The questions and answers on your GBP provide additional entity data to AI systems. When you answer them thoroughly, you’re creating structured information that AI systems can extract and use in recommendations.

IV. The NAP Consistency Problem That Kills AI Visibility

NAP consistency — your Name, Address, and Phone number being identical across every platform and directory — has always mattered for local SEO. For AI visibility, it’s even more critical.

AI systems cross-reference your information across multiple sources to verify your entity. Your website says one address. Your GBP says another because you moved two years ago and forgot to update Yelp. Your Facebook page has an old phone number. Your listing on the Chamber of Commerce website uses your DBA instead of your legal name.

To a human, these are minor inconsistencies. To an AI system synthesizing information to build an entity profile, these inconsistencies create uncertainty. Is this the same business? Is this business still operating? Which information is current? The AI can’t tell. So it recommends the competitor whose information is perfectly consistent across every platform.

This is one of the most fixable AI visibility problems and one of the most commonly ignored. Most local businesses have NAP inconsistencies across five to ten platforms and don’t know it. Old directory listings. Outdated chamber of commerce pages. Former addresses on industry-specific directories. A phone number that changed three years ago still showing up on Yelp.

Run a citation audit. Check every directory, every platform, every listing for your business. Make them all match. Exactly. Same name formatting. Same address. Same phone number. Same website URL. This takes a day of tedious work and it removes one of the most common reasons AI systems skip local businesses.

V. Schema Markup: The 3.2x Lever Local Businesses Ignore

Structured data is a 3.2x lever on AI citations. Businesses with comprehensive schema markup appear in AI recommendations at 3.2 times the rate of businesses without it. And the vast majority of local businesses have either no schema markup or only the most basic implementation.

For local businesses, the schema types that matter most are LocalBusiness schema with all sub-properties filled out. Service schema for each service you offer with descriptions and pricing information. Review schema that exposes your ratings and review count to AI systems. FAQ schema for common questions about your business. GeoCoordinates schema that pins your location precisely. OpeningHoursSpecification schema that tells AI systems when you’re available.

Most local business websites were built by web designers who either don’t know about schema markup or implemented a bare-minimum version years ago. The site looks great. The design is modern. The photos are professional. But the underlying code tells AI systems almost nothing about what the business actually does, where it’s located, what it charges, or whether it’s open right now.

This is why a simple site with comprehensive structured data might outperform a $200,000 custom build for AI visibility. The expensive site looks better to humans. The structured site is legible to machines. In a world where AI agents and AI search systems are increasingly mediating how customers find businesses, machine legibility is worth more than visual design.

The implementation cost is minimal. A competent web developer can add comprehensive schema markup to a local business website in a single day. The ongoing maintenance is negligible. And the visibility impact compounds for as long as the markup is live.

VI. The Local AI Visibility Audit

Here’s the audit we run for every local business we work with.

Start with the AI recommendation test. Ask ChatGPT, Perplexity, Google AI Mode, and Claude to recommend businesses in your category and your city. Do this for five variations of the query: “best [category] in [city],” “who should I hire for [service] in [city],” “recommend a [category] near [neighborhood],” “[category] [city] reviews,” and “affordable [category] in [city].” Document whether your business appears in any of the 20 resulting answers.

Check your review profile. Star rating across Google, Yelp, and Facebook. Review velocity over the last 90 days. Response rate to reviews. If you’re below 4.0 on any major platform, that’s priority one. If you haven’t responded to reviews in the last month, that’s priority two.

Audit your GBP completeness. Primary and secondary categories. Business description with entity-rich language. Services and products sections filled out. Photos updated in the last 30 days. Posts within the last two weeks. Q&A section with thorough answers. Hours current and accurate.

Run a NAP consistency check. Your name, address, and phone number on your website, GBP, Yelp, Facebook, industry directories, chamber of commerce, Better Business Bureau, and any other platform where you’re listed. Every inconsistency is a reason for an AI system to skip you.

Check your schema markup. View the source code of your website or use Google’s Rich Results Test. Is LocalBusiness schema present? Is it comprehensive? Does it include services, geo-coordinates, opening hours, and review data? If you see nothing, that’s the biggest quick win available.

Score yourself honestly on each of these five areas. The businesses that score well on all five are the 1.2% that get recommended. The businesses that score poorly on even two of them are likely in the invisible 98.8%.

VII. Why 45% Changes Everything

Forty-five percent of consumers now use AI tools to find local services. A year ago that number was 6%. That’s not a gradual shift. That’s a behavior change happening at the speed of ChatGPT adoption. And it’s accelerating as AI systems get better at local recommendations and as more people develop the habit of asking an AI instead of Googling.

For a local business, this means that roughly half of your potential customers are now being filtered through an AI system before they ever see your Google listing, your website, or your ads. If the AI recommends you, you get a pre-qualified visitor who arrives with high trust and high intent. If the AI doesn’t recommend you, that potential customer never even learns you exist.

The conversion rate difference is significant. Visitors who arrive after an AI recommendation convert at 1.2 to 5 times higher rates than traditional organic search traffic. Because the AI pre-qualified them. It didn’t just match a keyword. It evaluated whether your business was the right answer to their specific question. By the time they click through to your website, they’re already half-sold.

This means being in the 1.2% that AI recommends isn’t just a visibility advantage. It’s a conversion advantage. And the businesses that aren’t recommended are losing both visibility and the highest-converting traffic source available in 2026.

VIII. The Window Is Open But It’s Not Open Forever

88% of local businesses have no active strategy to appear in AI search results. That number is going to drop fast as awareness spreads. But right now, in July 2026, the vast majority of your competitors are doing nothing about AI visibility.

That’s a window. The businesses that audit their AI presence today, fix their NAP consistency this week, implement comprehensive schema markup this month, and start actively managing their reviews and GBP with AI visibility in mind, will have a compounding head start that late movers can’t quickly replicate.

Entity authority compounds. Every month of consistent signals builds on the previous months. The business that starts building AI visibility now will be six months ahead of the competitor that starts in January 2027. And six months of compounding entity signals in a market where 88% of businesses aren’t even trying is an enormous competitive advantage.

Your Google ranking doesn’t matter if half your customers are asking an AI instead. Your beautiful website doesn’t matter if the AI never sends anyone to it. Your ad spend doesn’t matter if the AI recommends your competitor to the person who would have clicked your ad.

The 1.2% are winning a disproportionate share of the highest-intent, highest-converting traffic available. The 98.8% don’t know the game is being played.

Now you know.