Forty-five percent of consumers now use an AI assistant to find a local service, up from six percent a year ago. In the same window, ChatGPT was measured recommending just 1.2 percent of all local business locations. Put those two numbers next to each other and you have the local-search story of 2026 in a sentence: people are asking AI which plumber, which restaurant, which estate agent to use, and for almost every business the answer it gives does not include them. The old game was ranking on a page of ten blue links. The new game is being the one name the assistant says out loud, and most local businesses have not noticed the rules changed.
I run a web and AI agency on Mallorca, so I watch this from the worst possible vantage point: a place whose entire economy runs on strangers deciding where to eat, stay, and buy. For years that decision started with a Google search and a map full of pins. Increasingly it starts with a question typed into ChatGPT or Gemini, and the response is not a list to browse. It is a short, confident recommendation of two or three places, and everything else may as well not exist.
The New Local Search Is a Single Answer
The behavioral shift underneath this is faster than anything I have seen in fifteen years of building websites. BrightLocal's 2026 survey put AI use for finding local services at 45 percent, a more than sevenfold jump in a single year. Google has not collapsed, it still holds around 90 percent of conventional search-engine share, but the experience changed underneath the market share. Roughly 68 percent of Google searches now end without a click to the open web, and AI Overviews, the summary box Google writes for you, appear in about 68 percent of local searches and sit above both the paid ads and the organic results.
The mechanical effect is brutal for anyone who relied on being findable. When an AI Overview appears, the click-through rate to the top organic result drops by 58 percent, measured by Ahrefs across 300,000 keywords. Pew Research, watching real browsing behavior, found people click a traditional link 8 percent of the time when a summary is present versus 15 percent when it is not. Google's newer AI Mode is more extreme still, with around 93 percent of those sessions ending in zero clicks. The search did not disappear. The list of ten options you used to scroll through did.
That is the part that should change how a local owner thinks. You are no longer competing for position four instead of position seven. You are competing to be one of the two or three names that exist at all inside a generated answer. It is closer to winning a recommendation from a knowledgeable friend than to ranking in a directory.
Most Local Businesses Are Simply Not in the Answer
Here is the uncomfortable measurement. SOCi's 2026 Local Visibility Index looked at more than 350,000 business locations and found ChatGPT recommends only about 1.2 percent of them when asked for a local option. Eighty-three percent of restaurants do not appear in AI local recommendations at all. This is not a gentle reshuffling of who ranks where. It is a near-total filter, and on the wrong side of it your business is invisible to a fast-growing share of the people who would have walked in.
Winner-take-most dynamics are not new on the web, but local used to be the exception. A maps result had room for a dozen nearby options, and proximity alone got you seen. The generated answer has no such mercy. It names a few and stops. For the business that gets named, AI discovery is a compounding gift. For everyone else it is a slow leak of customers who never knew the place existed, with no analytics dashboard showing the loss, because you cannot measure the searches where you were never mentioned.
The Surprise: ChatGPT Does Not Ask Google
This is the finding that reframes the whole problem, and most local owners have never heard it. When researchers reverse-engineered where ChatGPT's first local recommendations come from, the answer was not Google. Roughly 60 to 70 percent of the local businesses ChatGPT surfaces first are pulled from Foursquare's Places API. The assistant is leaning on a location dataset most business owners last thought about in 2014.
Sit with what that means. Your Google Business Profile, the thing every local-SEO guide told you to obsess over, is one input among several, and for the assistant a lot of people now ask first, it may not be the input that matters most. Your listing on Foursquare, Apple Maps, and Yelp, the consistency of your name, address, and category across all of them, is now a direct ranking signal for AI recommendations. The businesses that quietly kept those listings clean are getting recommended by a machine they never optimized for, and the ones who let them rot are paying for it without knowing why.
This is why I tell people that local AI visibility is not Google visibility with a new coat of paint. It runs on a different and wider set of data sources, and the work is partly old-fashioned listing hygiene that fell out of fashion right when it started to matter again.
Your Website Still Matters, But Differently
None of this retires your website. It changes the job the website does. An assistant that wants to recommend you has to be able to read you, and a startling amount of the web is now unreadable to it. Adobe found that 34 percent of product pages and a quarter of homepage and category pages are inaccessible to AI assistants, usually because the content is rendered by JavaScript that the AI crawler never executes. If the assistant cannot parse your hours, your prices, your services, you are not in the answer, no matter how good the page looks to a human.
The fix is the unglamorous discipline of making your facts machine-readable: server-rendered text instead of script-built content, complete LocalBusiness and FAQ structured data, and clear answer blocks that state a fact in one place instead of smearing it across five paragraphs. The deeper mechanics of why an engine cites one page and ignores a better-written one are worth understanding properly, and I wrote a full piece on how AI citations actually work for anyone who wants the pipeline rather than the checklist. The short version is that engines lift passages, not pages, and they reward content that is reachable, liftable, and corroborated everywhere else on the web.
Mallorca Is Already Being Sorted by Algorithm
The travel verticals are where this is moving fastest, which makes the Balearics an early and slightly unsettling test case. Fifty-six percent of US leisure travelers used AI for at least one trip in the past year, generative AI now makes up around a third of trip-research activity, and AI-driven travel traffic to booking sites grew nearly 200 percent year over year in May. Europe is a step behind the UK on the same curve, with Spain squarely in the markets being tracked.
It is no longer abstract here. In June a Balearic outlet reported a tourism study finding that AI agents now recommend specific island destinations by name, Sant Antoni on Ibiza, Ciutadella on Menorca, acting as an amplifier that pushes already-crowded places harder. Big operators are moving to be inside that funnel rather than downstream of it. IHG launched a ChatGPT app in early June to let travelers discover and compare its hotels directly in the assistant. Real estate is just as exposed, with AI Overviews showing up for as many as half of local-intent property searches, which is precisely the territory an AI-ready real estate site is built for.
There is a saving grace, and it is an important one. Around 51 percent of travelers who use AI still click through to a real website before deciding, and AI-referred travel visitors still convert about 28 percent worse than non-AI ones, because booking a holiday or a finca is a trust decision a person still wants to make as a person. The assistant increasingly owns discovery. The human still owns the close. That split is the whole strategy: you want to be in the answer when someone asks, and you want a site and a process good enough to win them once the AI hands them over.
What to Actually Do
The practical work divides into three layers, and only the first is urgent. Start with presence on the surfaces the assistants actually read. Claim and clean your Foursquare, Apple Maps, and Yelp listings alongside Google, make the name, address, phone, and category identical across all of them, and treat that consistency as the ranking signal it has quietly become. This is cheap, it is mostly a weekend of careful work, and it is the single highest-return thing a local business can do for AI visibility right now.
The second layer is making your own site answerable. Server-render the facts, add LocalBusiness and FAQ structured data, and write the key questions a customer would ask an assistant directly into pages with direct, liftable answers. Treating this as its own discipline rather than an afterthought to classic SEO is the core of what generative engine optimization actually is, and the businesses doing it now are accumulating an advantage while only about one in seven of their competitors even tracks whether AI mentions them. For the verticals where the stakes are highest, the gap between the prepared and the invisible is already wide.
The third layer is the forward bet, and it is optional for now: giving agents a real interface to your business, an API or an MCP server they can query for availability, pricing, and booking instead of scraping a cached page. For most local businesses that is a 2027 conversation. But the discovery work in the first two layers is a today problem, and it is the rare kind that rewards moving before your competitors do.
What This Means
Local discovery is bifurcating, quietly and without an announcement. A small set of businesses is becoming the answer the assistants give, and that position compounds, because being recommended makes you more visible which makes you more likely to be recommended again. Everyone else is sliding toward an invisibility they cannot see in any dashboard, losing the customers who used to find them by accident on a map. The cruel part is that the loss is silent. There is no notification that you were left out of an answer.
My prediction is that within a year, asking which businesses an AI recommends in your category will be as routine as checking your Google ranking was a decade ago, and a lot of owners will not like what they hear. The good news is that the filter is still loose enough to climb into. Most local businesses have not done the listing hygiene, have not made their site machine-readable, and have no idea Foursquare is back from the dead. That gap is the opportunity, and on an island that lives or dies on being chosen by people who have never been here, I would not wait for it to close.
Written by Matthias Meyer of StudioMeyer, a web and AI agency on Mallorca building MCP servers, agent fleets and AI products for small and mid-size businesses. This article was originally published on the StudioMeyer blog.