Most café owners think about SEO the way they thought about it five years ago: get a Google Business Profile, collect some reviews, maybe blog once a quarter. That approach is losing ground fast. AI Overviews now appear on more than 13% of all searches, up from 6.5% just two months prior in early 2025. Gartner expects traditional search volume to drop 25% this year as users shift to AI answer engines. When someone types "café with wifi near me" into ChatGPT or Perplexity, they're not getting a list of blue links. They're getting one answer. You want to be that answer.
This isn't about chasing algorithms. It's about understanding how AI systems decide which café to recommend, and then giving them exactly what they need to pick yours.
Why café searches are moving to AI platforms
The shift is already visible in how people phrase their searches. "Best espresso bar in [city]" used to land on Yelp roundups and local blog posts. Now it often surfaces directly in an AI-generated response that synthesizes reviews, menu details, and location data without the user ever clicking through to a website. That 34.5% average reduction in organic clicks on top results isn't a rounding error. It's a structural change.
For cafés specifically, this matters because so much of the discovery happens in the moment: someone walking through a neighborhood, someone on a video call who needs a spot with good wifi, someone who wants a pastry and a cortado before noon. These are high-intent, local, time-sensitive queries. AI platforms are built to answer exactly these kinds of questions, and they're drawing on a specific set of sources to do it.
The cafés that get cited by ChatGPT, Gemini, and Perplexity aren't necessarily the most famous or the longest-established. They're the ones whose information is complete, machine-readable, and consistent across the places AI systems actually check.
The schema markup your café website needs right now
Schema markup is how you talk directly to AI crawlers in a language they trust. Most café websites either skip it entirely or use the generic LocalBusiness type, which leaves specificity on the table. Use Restaurant schema instead. It's a more specific type, and AI systems favor it when answering food and drink queries.
The properties that matter most: name, address, geo coordinates, telephone, priceRange, openingHoursSpecification, and currenciesAccepted. These aren't optional extras. When someone asks Google AI Mode or ChatGPT "is there a café open before 7am near the train station," the systems that answer confidently are the ones with complete location and hours data in their markup.
Menu schema is where most cafés are leaving the most visibility behind. When someone asks "do you have oat milk" or "can I get lunch under $15," AI search engines look for Menu and MenuItem schema to answer those questions. An estimated 60% of independent restaurant websites still serve their menus as PDFs, which AI crawlers can't read at all. Converting your menu to structured HTML with Menu > MenuSection > MenuItem markup, and adding suitableForDiet properties for vegan, gluten-free, or other options, makes your café matchable for a whole category of queries that PDF menus simply miss.
Beyond menu and restaurant schema, implement AggregateRating (pulling from your review data), FAQPage for common questions about parking, wifi, or reservation policy, and a ReserveAction if you take bookings. FAQPage schema in particular aligns closely with how AI platforms structure their answers, making your content easier to extract and cite.
The directories that AI systems actually use
Google Business Profile is non-negotiable, and that's been true for years. What's changed is the reason: Google AI Mode consistently uses GBP as a primary data source when generating local recommendations. A GBP with incomplete hours, a missing phone number, or outdated photos isn't just less appealing to human searchers. It's less likely to be cited at all.
Foursquare is the one that surprises most café owners. It has over 100 million points of interest across more than 200 countries, and ChatGPT pulls from it directly. If your café isn't claimed and filled out on Foursquare, you're invisible to a significant portion of ChatGPT's local data. MapQuest is another that both Google AI Mode and Perplexity reference, even though most people haven't thought about MapQuest since 2009.
For social platforms: Instagram was cited as a source by both Google AI Mode and Perplexity. Facebook showed up in Google AI Mode and ChatGPT results. This means your Instagram grid and your Facebook business page are functioning as AI citations, not just social content. Keeping them current with accurate hours, location tags, and photos of your specialty coffee and food matters more than it used to.
The practical checklist for citation sources:
- Google Business Profile (complete every field, add weekly photos)
- Foursquare (claim and verify your listing)
- MapQuest (often overlooked, genuinely matters for AI results)
- Yelp (still a major review aggregator for AI systems)
- Instagram and Facebook business pages
- Apple Maps (critical for Siri and Apple Intelligence queries)
How to write content that AI platforms actually cite
Perplexity attaches sources to its answers. It's asking, implicitly, whether your content is worth citing. That means your website needs to do more than describe the vibe. It needs to answer specific questions with specific facts.
A page that says "we serve great coffee in a cozy space" is not citable. A page that says "we serve single-origin pour-over and espresso drinks Monday through Saturday from 6:30am, with free wifi and outlets at every table" gives an AI system something to extract and repeat. That's the difference.
Gemini specifically favors pages that are logically structured with clear headings and direct answers. If someone searches "café with wifi and outdoor seating in [neighborhood]," and your website has a page or section with a heading like "Amenities and seating at [Café Name]" followed by a clear list of what you offer, Gemini can find it, parse it, and include it in an Overview. Bury that information in a paragraph on your About page and it might as well not exist.
For your content structure, numbered lists work well for things like ordering process or parking instructions. Bullet lists work for amenities. Short FAQ sections on your main pages, marked up with FAQPage schema, give AI platforms a ready-made citation format. Here's a simple content framework that works:
- Lead with your core offering (specialty coffee, pastries, lunch service) and your location with cross-street reference
- State your hours explicitly, including any seasonal changes
- List your amenities (wifi speed if you know it, number of seats, outdoor availability)
- Answer the top three questions customers ask before visiting
- Include prices or a price range for key items
- Link to your structured menu page, not a PDF
Making your café visible to different AI systems
ChatGPT, Gemini, and Perplexity don't all rank the same signals the same way. ChatGPT weighs authority, meaning it looks at whether reputable sources mention you. Local press coverage, being listed in a "best coffee shops" roundup from a credible outlet, or getting cited on a food blog with real readership all help here.
Gemini is more concerned with clarity and structure. If your site is easy for a machine to parse, logically organized, and uses schema markup, you're in good shape with Gemini. It favors content that maps directly to specific intent rather than general brand storytelling.
Perplexity wants verification. It checks whether your facts are consistent across sources. If your hours on your website say 7am but your GBP says 8am and your Foursquare says 7:30am, Perplexity loses confidence in your data and may not cite you. Consistency across every listing is more important than any single optimization.
This is also where your robots.txt file becomes useful beyond just search crawlers. Placing a plain-text description of your café at your domain root, telling AI crawlers which pages to prioritize (your menu, your contact page, your about page), and ensuring your site loads quickly all reduce friction for AI systems trying to understand what your café is.
Getting your café cited, not just ranked
The goal has shifted. Ranking on page one used to be the win. Now the win is being the café that an AI confidently recommends when someone asks for a great espresso bar with outdoor seating and good lunch options. That recommendation depends on complete data, consistent listings, readable content, and schema markup that tells AI systems exactly what you offer.
If you're not sure where your café currently stands across these signals, SuggestedByGPT runs a free scan that shows how visible your business is to AI platforms and where the gaps are. Most cafés find at least two or three significant issues they didn't know existed, usually around menu schema or inconsistent directory data.
These are solvable problems. They just require doing them deliberately rather than hoping your current website handles it.
Run a free AI visibility scan for your café at /start. You'll see exactly which signals are missing and what fixing them would mean for how AI platforms describe and recommend your business.