The AI Recommendation Gap: Why Almost No Local Businesses Get Named

Your customers are already asking ChatGPT where to eat, which plumber to call, and who fixes a gas furnace at 11pm. According to recent industry research, 45% of consumers used AI tools to find local businesses in the past year — up from just 6% the year before, making AI the third most-used local discovery channel behind Google and Facebook.

Here's the uncomfortable part: almost none of those searches name an actual business. SOCi's analysis of more than 350,000 locations across 2,751 brands found that ChatGPT recommends only 1.2% of local business locations, compared with a 35.9% average appearance rate in Google's local 3-pack. In other words, getting recommended by AI is 3 to 30 times harder than ranking in traditional local search.

That gap creates a strange new reality for local operators:

• 31% of consumers have asked ChatGPT specifically for business recommendations, and 23% have used Google AI Mode.
• 63% of AI users trust the recommendations they get — though 88% verify them afterward against reviews.
• Roughly 99% of local locations are never named at all, in most categories and most cities.

Before you panic about that 1.2% figure, understand what it actually means. As the research puts it: if the AI never names you, it almost certainly isn't naming most of your competitors either — which means the slot is still open. AI recommendations aren't a settled ranking hierarchy the way the Google 3-pack is; they're a sparsely populated field where the winner often wins by default, simply by being the business the model can verify and cite.

That reframes the problem entirely. This isn't about displacing an entrenched incumbent. It's about showing up at all in a channel where almost nobody has. And because AI search referral traffic grew 527% between January and May 2025, the businesses that close that gap early are capturing demand their competitors don't even know exists.

The catch is that you can't fix what you don't measure. As one practitioner bluntly put it, you cannot manage a channel you have never looked at. Most local businesses have never once checked whether ChatGPT, Gemini, or Perplexity mentions them — a blind spot that tools like Briefsy are built to close, by running AI-search visibility checks across the major engines and flagging exactly where your name shows up, gets mangled, or doesn't appear at all.

The good news buried in the 1.2% number: the bar is low, the field is empty, and the sources that feed AI recommendations are largely things you already control. The next question is figuring out which sources those actually are — and the research disagrees more than you'd expect.

How ChatGPT Actually Picks Businesses (and Why Google Rankings Don't Guarantee It)

Here's what most local business owners don't realize: ChatGPT has no local ranking formula at all. When it answers "where should I eat tonight?" or "who can fix my furnace?", it runs a web search, pulls from its training data plus live integrations like Bing, and assembles an answer from whatever sources it finds credible at that moment. There's no published scoring system, no paid placement option, and — contrary to a common assumption — no requirement that a business even has a Google Business Profile for ChatGPT to mention it.

The mechanics matter for anyone trying to appear in these answers. OpenAI's crawler, OAI-SearchBot, is what determines whether your site can show up in ChatGPT Search results — and blocking GPTBot alone doesn't keep you out, while opting out of OAI-SearchBot does. Beyond crawling, ChatGPT doesn't accept paid placements for local recommendations; visibility depends on data accuracy, relevance, and source authority, not budget.

But here's where the research gets genuinely confusing. Different audits tell different stories about which sources ChatGPT actually trusts:

• A 6.8M-citation analysis found 86% of AI citations come from business-controlled sources — 44% from the business's own website and 42% from listings and directories.
• A 309-answer dining audit found the opposite: editorial content appeared in 48.9% of the sentences naming businesses, while listing pages showed up 0% of the time — though the factual "card" details often came from a licensed Yelp data feed.
• Practitioner analysis points to Bing Places for Business as ChatGPT's biggest data source, with no evidence it uses Google Business Profile data at all.

The most likely explanation is that these audits measured different things — citation types, business categories, query phrasing. But the practical takeaway is clear: don't assume your Google rankings or a polished GBP automatically carry into AI answers. The data cuts both ways — only 38% of Google AI Overview citations now go to top-ten-ranked pages, even though 76.1% of those citations also rank in Google's top 10. Ranking well helps, but it's not a guarantee.

There's one more wrinkle worth knowing before you test this yourself: ChatGPT's answers are unstable. In one audit, asking the identical question 100 times produced the exact same sources only 8 times. A single screenshot of ChatGPT recommending (or ignoring) your business is essentially noise. That's why we built Briefsy around repeated, structured AI-search visibility checks across ChatGPT, Gemini, and Perplexity — one answer at one moment tells you almost nothing, but a tracked pattern over weeks tells you whether you're actually in the conversation or still on the outside of it.

What Actually Moves the Needle: Specificity, Source Accuracy, and Review Detail

The businesses winning AI recommendations aren't the biggest — they're the most specific. When one agency tracked roughly 400 prompts across about 50 businesses, a water-heater-only company outranked a higher-authority general plumber, according to the same audit's findings.

Review quality follows the same pattern. A business with 2,000 short reviews was repeatedly skipped, while a competitor with just 50 detailed reviews kept getting recommended. As one agency owner put it, businesses whose homepage says "quality solutions for all your needs" basically don't exist to these tools.

The reason is structural. Local recommendations are question-specific: a query like "Who can repair a gas furnace in Aurora tonight and works on rental properties?" contains four distinct requirements, each needing verifiable business information. AI models can't confirm what you never stated explicitly.

So how do you act on this? Match the fix to your current situation:

• Absent from answers: close source gaps first — complete Bing Places, make your NAP consistent across directories, and publish answer-shaped content that states services, hours, and limits directly.
• Mentioned but wrong: fix stale data at the source. AI answers typically update within weeks once the underlying sources are corrected.
• Mentioned accurately: maintain review velocity and content freshness, and re-check monthly — identical questions produced identical sources only 8 times in 100, so single screenshots prove nothing.

Remember that 86% of AI citations come from sources businesses already control: 44% from your own website and 42% from listings and directories. There is no submission form — visibility is earned at the sources.

The practical challenge is that these signals drift constantly, which is why monitoring matters. Tools like Briefsy exist to watch this layer for you — tracking AI visibility, reviews, and competitor changes so you know which of the three scenarios you're in this month, not last quarter.

Whichever way you check, the framework holds: diagnose your status, fix at the source, and measure again. That loop, run consistently, is what turns a one-off recommendation into a repeatable channel.

Your 10-Minute AI Visibility Check (and How to Make It a Repeatable System)

A single ChatGPT screenshot tells you almost nothing about whether AI tools actually recommend your business. Identical questions produced identical sources only 8 times in 100 in one audit of 309 AI answers, which means any one-off check is closer to a coin flip than a measurement.

The fix is structure, not luck. Set aside about 10 minutes each month and run the same three prompt types across ChatGPT, Gemini, Perplexity, and Google AI Mode:

• Recommendation prompts — ask as a customer would: "Who's the best plumber in [city] for emergency repairs?"
• Brand prompts — ask what the AI knows about your business specifically, to catch mentioned-but-wrong answers like stale hours or services.
• Decision prompts — use hyper-specific queries with multiple requirements, since that's how real customers phrase questions when they're ready to buy.

Three rules make the results meaningful. First, use fixed questions every month so you're comparing trends, not random draws. Second, run each question twice to smooth out volatility. Third, never include your own business name in recommendation prompts — you want to see whether the AI names you unprompted, the way it would for a real customer.

Treat each result as "an answer to one specific question at one point in time," not a permanent ranking, as the ShiftFlow playbook puts it. Then sort findings into three outcomes: absent (close source gaps like incomplete listings and inconsistent NAP data), mentioned-but-wrong (fix stale data at the source — answers usually update within weeks), or mentioned-accurately (maintain review velocity and fresh content).

This is where manual checks hit their ceiling. Four platforms, three prompt types, two runs each, every month — plus watching whether competitors appear in answers you don't — turns a 10-minute task into an ongoing operational burden. That's the loop Briefsy automates: AI-search visibility checks across ChatGPT, Perplexity, and Gemini, competitor tracking, and a prioritized brief of what changed and what to do next, so AI visibility gets managed like any other channel rather than checked once and forgotten.

As one industry analysis bluntly observes, "you cannot manage a channel you have never looked at." Ten structured minutes a month is the minimum price of admission — automation just makes it sustainable.