AI visibility for cafés
The oat milk, the outlets, the dog under the table —that's what AI is actually readingbefore it decides which café to name
Nobody asks ChatGPT for "a café." They ask for a café that's open right now, has oat milk, allows dogs, or has a table free for a 2pm work session. If your listing doesn't answer the specific question, the AI moves to the café that does — even if yours pulls the better crowd on weekends.
The person asking isn't browsing — they're standing on a corner deciding where to walk in the next three minutes. If the AI isn't sure you're open, have the milk they need, or have a seat free, it names someone else.
Cafés live and die on small, changeable facts: hours that shift on Sundays, a machine that's down, a terrace that closes when it rains, a "no laptops after 12" rule during lunch rush. An AI assistant answering a café question is stitching together your Google Business Profile, your website, review text, and whatever aggregator last touched your hours — and when those disagree, most assistants default to caution or silence rather than risk sending someone to a closed door. That's different from a restaurant reservation or a retail purchase: a café recommendation is almost always acted on within minutes, on foot, so the AI is optimizing for "won't get this person turned away" over "best possible match." Menu detail matters more than most owners assume, too — a customer asking about oat milk, decaf, or gluten-free pastries wants a yes/no answer, and if your menu lives only as a photographed PDF, there's nothing for the AI to read.
How your customers ask
How customers actually ask
Why it matters
Where cafés actually lose the answer
Sunday hours are the ones that lie
Most café hour changes happen on the margins — a later Sunday open, an early holiday close, a one-off closure for a private event. Google Business Profile, your website, and Yelp rarely get updated in the same hour, and AI assistants that can't resolve a conflict tend to hedge or skip you rather than guess wrong and send someone to a locked door.
The laptop crowd is a policy question, not a vibe
"Good for working" is one of the most common café questions an AI gets asked, and it needs a real answer: outlet count, a laptop cutoff at lunch, wifi that actually holds up, whether two hours nursing one coffee is welcome. If that policy only exists as something your staff explains at the counter, the AI has nothing to cite and defaults to a competitor whose website spells it out.
Milk, allergens, and dietary swaps live or die on your menu format
A customer asking about oat milk, gluten-free pastries, or nut allergies wants a definite answer before they walk over, not a call-ahead recommendation. If your menu is a photo of a chalkboard or a PDF, an AI assistant can't parse it — the same information written as plain text on your site is far more likely to get surfaced.
Go deeper
Articles for Cafés
Fundamentals
AI visibility for cafés: why ChatGPT does (not) recommend your coffee house
Practice
Vegan, gluten-free, oat milk: how the AI understands your café menu
Data & studies
Ask the AI about your café: how to test your visibility yourself
Strategy
The laptop guest as a target group: how to show up in AI searches for work cafés
Data & studies
Opening hours, dogs, terrace: why contradictory data kicks your café out of AI answers
Questions café owners actually ask
We already keep Google Business Profile updated — isn't that enough?
It's necessary but not sufficient. GBP is one of several sources an assistant cross-checks, and if your website, delivery apps, or old directory listings still show a different set of hours or an outdated menu, the assistant sees a conflict, not just an out-of-date field on GBP. The fix is consistency across every place your info lives, not perfection on one.
Do we need special AI markup or an llms.txt file to show up?
No. Google has said directly that no special markup or AI-specific files are required for AI Overviews or AI Mode, and independent testing has found llms.txt files aren't read by Google Search at all and rarely drive measurable traffic anywhere else. Plain, accurate, consistent text on your existing site does more than any AI-only file.
Is there a food-safety or health-claim risk in how we describe our menu for AI?
Yes, and it's worth being precise about. If you state "gluten-free" or "nut-free" for an item, that's a food-safety claim customers with allergies will act on — write it only where it's actually accurate and kept current, since an AI assistant will repeat it as fact with no way to know your kitchen changed a recipe. Vague marketing language is safer than a wrong specific claim.