Ask ChatGPT to recommend a local business and there is a 98.8% chance your business will not be one of the options it gives back. That is not a typo. SOCi's 2026 Local Visibility Index found ChatGPT surfaces just 1.2% of local businesses when asked for a recommendation in a given category. For restaurants specifically, 83% do not appear in AI local recommendations at all, regardless of how good the food is or how many five star reviews sit on Google.
Meanwhile, the number of people asking in the first place is exploding. BrightLocal's 2026 Local Consumer Review Survey found 45% of UK consumers have used a tool like ChatGPT, Gemini or Perplexity to find a local business in the past year, up from just 6% in 2025. More people are asking AI where to go. Far fewer businesses are getting mentioned when they do.
1. The recommendation gap
Not all AI tools are equally stingy. Gemini, which draws on Google Maps data, recommended 11% of local businesses in the SOCi study. Perplexity, which blends web search with citations, managed 7.4%. ChatGPT trailed both at 1.2%. For comparison, brands still appeared in Google's own local 3-pack 35.9% of the time, a reminder of just how far AI chat tools have to go before they match traditional local search.
| Channel | Share of Local Businesses Recommended | Notes |
|---|---|---|
| Google local 3-pack | 35.9% | Established ranking system, high coverage |
| Gemini | 11% | Grounded in Google Maps data |
| Perplexity | 7.4% | Blends web search with citations |
| ChatGPT | 1.2% | Lowest recommendation rate of the four |
Local AI visibility researchers now describe the gap as up to 30 times harder to close than ranking in Google's local pack. The businesses that do get mentioned are not necessarily the best ones. They are the ones with the clearest, most consistent digital footprint for the model to draw on.
2. When AI does mention you, it might be wrong
Getting mentioned is only half the problem. The BrightLocal research also tested accuracy once a business was named, and the results are uncomfortable reading for small business owners specifically. Business profile information was only about 68% accurate on ChatGPT and Perplexity, compared with close to 100% on Gemini, which benefits from being grounded directly in Google Maps rather than open web text.
The gap widens further once you split results by business size. AI chatbots confused or misattributed SME brand names five times more often than large company names, at 4% versus 0.7%. They described an SME's services accurately only 87% of the time, against 93% for large brands. Worse, they confidently fabricated facts about small businesses at more than double the rate they did for large ones, 5% versus 2%, and were more likely to have no information at all: 6.3% of key details missing for SMEs compared with 4.8% for large brands. The categories AI got wrong most often were exactly the ones customers rely on: business size, contact details and how long the business has been trading.
| Accuracy Measure | SME Businesses | Large Brands |
|---|---|---|
| Business info accuracy (ChatGPT/Perplexity) | 68% | 68% |
| Name confused or misattributed | 4% | 0.7% |
| Services described accurately | 87% | 93% |
| Confidently fabricated facts | 5% | 2% |
| Key information missing entirely | 6.3% | 4.8% |
3. Why small businesses lose twice
Put the two findings together and a small business faces a double bind. It is far less likely to be recommended by AI in the first place, and on the rare occasion it is mentioned, it is more likely to be misdescribed, confused with another business, or missing basic facts entirely. Large brands do not escape AI inaccuracy altogether, but they have more structured data across more places for a model to draw on, which narrows the gap considerably.
This matters because, for a growing share of consumers, the AI's answer is the only answer they see. There is no scrolling past it to check a second opinion. If the model gets your opening hours wrong, or attributes a competitor's speciality to you, or simply never mentions you, that is the version of your business a customer walks away with.
4. What actually feeds the machine
AI models build their answers from whatever structured, consistent information about a business exists across the web, not from a single source. A few things reliably help:
- Keep your name, address, phone number and opening hours identical across every platform, Google, Facebook, directories and your own website. Inconsistency is a leading cause of confusion and missing data.
- Treat your Google Business Profile as a data anchor, not just a listing. Gemini's near-100% accuracy came specifically from being grounded in Google Maps data, so a complete, current profile feeds more than just Google's own search results.
- Write a clear, specific description of what you actually do. Vague or generic service descriptions give a model more room to guess, and guessing is where fabrication creeps in.
- Keep review content detailed rather than one-line. AI models draw on review text itself, not just star ratings, to describe what a business offers.
- Check periodically what ChatGPT, Gemini and Perplexity actually say when asked about your business. It is the only way to catch a wrong answer before a customer does.
The gap is a warning, not a curiosity
Forty five percent of consumers already ask AI where to go before they ask anyone else. That number was 6% a year ago. A recommendation rate under 2% and an accuracy rate that gets worse the smaller you are is not a quirky statistic to file away. It is a visibility gap that is widening while consumer behaviour shifts underneath it. The businesses that close that gap will not be the ones who wait for AI search to mature. They will be the ones who make sure the web already tells a consistent, accurate story before anyone asks a chatbot to repeat it.