AI assistants are another route through which people discover local providers. Two studies show why that route merits attention. They describe different samples: BrightLocal surveys US consumers, while SOCi examines multi-location brands. Their results are not a census of every business on Mallorca.
What the consumer survey actually shows#
In BrightLocal’s Local Consumer Review Survey 2026, 45 percent of respondents said they had used AI for local business recommendations in the previous twelve months, up from six percent a year earlier. The panel consisted of 1,002 US adults. This is annual usage, not the share of all local searches or a finding about Spain.
The useful business question is narrower: do your customers ask an assistant, and what answer do they receive? Google does not need to be replaced by a new channel. You can observe additional routes while maintaining the same reliable business facts across relevant surfaces.
What the visibility study measures#
SOCi’s Local Visibility Index 2026 examined more than 350,000 locations belonging to 2,751 multi-location brands. In that research setting, ChatGPT recommended 1.2 percent of locations. Gemini and Perplexity produced different figures. This is a result for those brands, questions and measurement conditions, not proof that 98.8 percent of every local business worldwide is invisible.
It nevertheless highlights a useful distinction: appearing in conventional local search and being mentioned in a particular AI answer are different outcomes. A missing mention in one test does not mean the provider is never recommended.
Sources vary by system and question#
Assistants may use search results, map data, business profiles and websites. Sources for a particular answer depend on the product, question and search mode. We cannot infer a universal Foursquare share or a fixed ranking signal across every provider.
Start with profiles relevant to your business and customers: Google Business Profile, Apple Maps where appropriate, industry directories and other map services. Align the name, address, telephone number, category, hours and website. Inspect source links in actual AI answers before prioritising another service. This is a practical approach, not a guarantee of a recommendation slot.
Your website remains a verifiable source#
Customers should quickly understand services, location, prices or pricing conditions and contact options. Machines should be able to retrieve the same facts. Inspect delivered HTML and response codes rather than inferring readability from a framework name. Important facts should be accessible without later JavaScript when your target crawler does not execute it.
Suitable LocalBusiness markup can structure visible facts. Markup does not replace content or accuracy. Google explicitly says AI Overviews and AI Mode require no special AI files or additional schema types. Llms.txt can help compatible clients as an overview, but is not a universal citation requirement. Our approach to the website itself is described under AI visibility.
What this means for Mallorca businesses#
Visitors and island residents ask different questions about distance, languages, opening hours, availability and particular services. Collect questions that genuinely precede a booking. Check multiple languages where you can serve those customers. An accurate Spanish description and clear German service page help more than a blanket claim to be optimised for every AI.
An AI mention does not complete a decision. Your website must make the recommendation verifiable and provide a working next step: contact, appointment, enquiry or booking. Measure the resulting enquiries and keep mentions separate from visits and completed transactions.
Three practical layers of work#
First, check business facts. Correct inconsistent names, old addresses and wrong hours on important profiles. Record where information appears and who maintains it. You do not need every conceivable directory.
Second, improve answers on your website. Explain services, coverage and contact clearly. Use your own photographs and verifiable facts. Put genuinely different needs on suitable pages rather than generating many near-identical location or keyword pages. Check crawler access, internal links and actual content delivery.
Third, measure observed answers. Use open questions without inserting the business name. Record the question, language, region, date, system and search mode, and repeat the queries. Keep answers and their sources. A test is a sample, not a complete market view. Bing Webmaster Tools adds citation data across supported Microsoft surfaces; a citation is not a click.
Interfaces need a concrete use case#
An API or MCP server can make sense when a connected assistant must check availability, prices or bookings. It needs an actual client, defined permissions, current data and controlled actions. Publishing an interface alone does not make general AI search discover or use it.
For many businesses, the first task is straightforward: correct facts, answer important customer questions and offer a reliable enquiry process. Any additional AI visibility is then observed. The work should help your business instead of selling an untestable ranking guarantee.
