AI Local Search Has a Visibility Problem: 85.6% of Venues Never Appeared in Recommendations

One of the most interesting findings is the distinction between being selected and being ranked highly once selected. Businesses might assume that having the highest star rating is the obvious route into an AI recommendation, but the study found something more complicated. At the initial selection stage, star rating was not significantly associated with whether a venue entered the recommendation set. Instead, signals connected to how well documented the business was online showed stronger associations. Having an owned website was associated with 1.92 times the odds of entering an AI recommendation, while review volume, listed pricing information and third-party web mentions were also positively associated with inclusion.

This matters because AI visibility is not simply a digital popularity contest. A business may be excellent offline and have loyal customers while still giving AI systems relatively little information to work with. Another venue may have a clear website, detailed service information, accessible pricing, numerous independent references and consistent information across the web. That creates a richer body of evidence for an AI system trying to decide whether the business matches a user's request. The implication is not that any one of these factors guarantees visibility, but that businesses which are easier to understand and verify online may have an advantage when AI systems build recommendation sets.

The finding around owned websites is particularly important because some local businesses have spent years treating Instagram, Google Business Profile or booking platforms as substitutes for maintaining a strong website. That approach becomes increasingly risky as AI-powered discovery grows. A well-maintained website can establish what the organisation is, where it operates, what it offers, what makes it distinctive, its opening hours, pricing, menus, services, contact information and other details that could influence a recommendation. The website is therefore no longer only somewhere customers visit after discovering a business. It can also form part of the evidence machines use to understand it.

Clear information becomes even more important when AI searches become specific. A customer might not ask for the “best restaurants in Leeds”. They could ask for somewhere suitable for six people where mains are around £20 and vegetarian options are available, or somewhere near the station that is open on a Monday and appropriate for a client lunch. The more clearly a business documents those details, the easier it becomes for an AI system to evaluate whether it matches the request. Marketing teams therefore need to think beyond broad category keywords and consider whether important commercial facts are explicit, accessible and current.

Third-party information strengthens that picture further. The study found an association between external web mentions and entering recommendations, reinforcing the idea that what a company says about itself is only one part of its digital identity. Editorial articles, local press, review platforms, directories, recommendation pages, industry websites and other independent sources can all contribute information about a business. This is where AI visibility begins to overlap with digital PR, online reputation management and local SEO. The objective should not be to manufacture low-quality mentions, but to develop a credible and consistent presence across the places where customers and authoritative third parties naturally discuss the category.

Star ratings became more significant once businesses had already entered the recommendation set. Although rating did not significantly predict whether a venue was selected initially, it was associated with being placed first among venues that had already been recommended. That creates a useful distinction for marketers: first, the business needs to become part of the consideration set, then it needs to compete for prominence within that answer. Strong reviews remain valuable, but they may not solve an underlying discoverability problem if AI systems still have insufficient information about the business itself.

The research also revealed an accuracy issue. Outright fabricated venues were relatively rare, but permanently closed businesses were recommended on 93 occasions. That suggests stale information can be a more practical problem than complete hallucination. Opening hours, locations, menus, pricing, service information and business status change frequently, and outdated information can remain scattered across the web long after a business updates its own website. For marketers, maintaining accurate digital information is therefore becoming part of AI visibility management. Being recommended is not particularly valuable if the recommendation contains information that frustrates or misleads the customer.

Another reason businesses should avoid treating one ChatGPT search as a complete AI visibility audit is that the four systems did not recommend identical businesses. The researchers found relatively low cross-system agreement, with recommendation overlap varying considerably between ChatGPT, Claude, Gemini and Perplexity. They also found variation between repeated runs. A business may therefore appear frequently in one platform and barely appear in another, while repeating the same question may produce a somewhat different shortlist.

This makes AI visibility a measurement problem as much as an optimisation problem. Testing one prompt once provides little more than a snapshot. A more useful approach is to identify the questions potential customers are likely to ask, test them across the AI platforms that matter, record which competitors are appearing and monitor whether those patterns change over time. Businesses also need to assess whether they are simply mentioned, strongly recommended or accurately represented, because those are different outcomes.

The study does not provide a formula that guarantees inclusion in every AI recommendation, but it does highlight several areas marketers can control. Businesses can improve the quality and clarity of their owned website, make important commercial information explicit, keep details consistent and current, develop legitimate reviews and build a stronger third-party presence around the brand. Just as importantly, they can begin measuring what AI systems actually say rather than assuming traditional search performance automatically translates into AI visibility.

Local SEO is not disappearing. Many of the fundamentals that make a business easy to discover, understand and trust remain valuable. What is changing is the interface between the customer and the information available online. Sometimes the customer will use Google Maps, sometimes they will search Google, and increasingly they may ask an AI assistant to choose on their behalf. When that happens, simply having a good business is not enough. The business needs to become part of the answer.

Primary research referenced in this article is Invisible to the Machine: Auditing AI Restaurant, Cafe, and Bar Recommendation Against a Complete Market Census, by Vladimir Pitenin, published on 7 August 2026. The study analysed 4,776 venues across Canggu and Ubud and compared them against recommendations generated by ChatGPT, Claude, Gemini and Perplexity.