GEO for specialty coffee shops
Coffee has the most attribute-driven queries in hospitality and one of the strongest specialist ecosystems. Both work in favour of a small café that publishes properly.
Short answer
Someone asking an assistant for coffee is almost never asking only for coffee. They are asking for a filter roast, or a power socket, or somewhere that will not mind a laptop for two hours, or a place open at seven on a Sunday. Cafés that publish those details in crawlable text win those queries. Most publish a hero image and an Instagram feed.
What we found when we checked
We audited eight well-known Melbourne specialty coffee businesses in September 2026, fetching each homepage the way an AI crawler does. The results were not what the industry assumes.
None of them were blocked. None were empty JavaScript shells. Every one served real text. The failure was somewhere else entirely: five of the eight had no usable H1 text, three had no <h1> element at all and two wrapped only a logo image, and only one of the eight stated its name, street address and trading hours in the HTML. None of the eight carried LocalBusiness or CafeOrCoffeeShop structured data.
The full method and the quoted evidence is in the teardown. The short version: this category is not losing to a technical block. It is losing to opening paragraphs that say nothing a query can match.
The attributes that decide coffee queries
- Who roasts it. Your own roast, or whose. Specialty queries name roasters constantly.
- Brew methods available: espresso, batch brew, pour-over, and which devices. “Filter coffee near me” is a real and frequent query.
- Current origins on the bar, if you rotate. This is the query specialty drinkers actually run.
- Wifi, and whether there are power sockets. Say the number if it is small.
- Laptop policy, explicitly. If laptops are fine until 11am and not after, publish exactly that. It wins both queries.
- Seating: how many, indoor and out, whether there is a communal table.
- Food, and how substantial. “Coffee and a proper breakfast” is a different query from “coffee”.
- Milk alternatives, named. Oat, soy, almond, and whether there is a surcharge.
- Opening hours including weekends, which is where most café queries actually land.
- Dogs, prams, step-free entry. Cheap to state, frequently decisive.
The opening paragraph problem
Nearly every café site opens with mood. The stored snippet an assistant keeps for your page is roughly 200 characters taken from around your H1, and it is the same extract for every question. “Where good coffee lives” answers nothing.
Replace it with something like: “A twenty-seat specialty café on Gertrude Street, Fitzroy. Our own filter roast, batch brew and pour-over, oat and soy at no surcharge. Open 7am to 3pm daily, laptops welcome before 11.” That sentence answers six query families and still sounds like a person wrote it.
The specialist ecosystem is an advantage
Coffee has an unusually strong network of specialist publications, city guides and dedicated review communities, and assistants cite them. For a small café this is better news than it is for most categories: the sources that matter are reachable, human-scale and interested in exactly what you do, which is a far easier outreach problem than national press.
Off-site mentions track AI visibility roughly three times more strongly than backlinks, on correlational evidence we are careful to label as correlational. For cafés it is also the most achievable half of the work.
Four disciplines, in the order the evidence supports
Common questions
No, and this is the most common situation in the category. Instagram is not crawlable by the answering bots in any useful way, so a business that lives there has almost nothing for a model to retrieve. A single, well-written page that states what you are, where you are, what you roast and when you open does more for AI visibility than a year of posts.
It is one of the most-asked constraints in the category and almost nobody publishes it. Stating it plainly wins you the people who want to sit and work and, just as usefully, the people who want somewhere that is not full of laptops. Both are queries.
Add it, and do not expect it to change your AI visibility. A matched study found adding JSON-LD produced no major uplift in citations on any platform. Schema earns conventional rich results and helps disambiguate you from a similarly named business, which are real reasons. It is not the lever it is sold as.
Find out which half is your problem
Book a 30 minute call. We will fetch your site the way the crawlers do, check whether the answering bots can reach it, and look at how the assistants answer your category today.