Cafés and restaurants are losing bookings to an answer engine, not to Google
Hospitality lives in the map pack and, increasingly, in whatever an assistant says when somebody asks for lunch nearby. Here is how to find the venues that are invisible to both.
Hospitality has a specific and unusually recent problem. The trade has always lived in the map pack, and most venues have made some peace with that. What has changed is that a growing share of "where should we eat" now goes to an assistant instead of a search box, and the signals an assistant needs are not the ones a restaurant's marketing person has been optimising.
The result is a category full of venues that look fine on Google and cannot be recommended by anything else.
Where the money is hiding
Opening hours that only exist in an image. This is the single most common finding in the trade and the one owners find most surprising. Hours live in a designed graphic, or in a PDF menu, or on a third-party booking widget that renders after the page loads. A crawler sees none of it. An assistant asked "is this place open on Sunday" cannot answer, so it recommends somewhere that can.
No entity definition. Without Restaurant or LocalBusiness schema an AI engine cannot confidently resolve the venue as a thing: which cuisine, which street, which of the three places with a similar name. It falls back to citing a directory listing or an aggregator instead of the venue's own site, and the venue pays that aggregator for the booking.
Thin homepage content. Hospitality sites are visual by nature and often carry almost no text. A page with a hero image and eleven words cannot rank for anything beyond the venue's exact name, and cannot be quoted by anything at all.
Too few reviews. Same mechanic as any local trade, sharper here: in a dense high street the review count is the tiebreaker between two places a hundred metres apart.
What the audit gives you that a screenshot does not
An AI-visibility conversation is easy to have badly. "You need an AI strategy" is not a sentence a café owner will act on, and it is not a claim you can defend.
What is defensible is what a SiteAssay report actually contains: which named crawlers the site's robots.txt blocks, whether valid structured data is present, whether the opening hours exist anywhere a machine can read them, and what a competitor down the road has that this venue does not. Those are checkable facts with obvious fixes, and they are cheap fixes, which makes them a good first job rather than a project.
Running the sweep
In the search screen, pick the Cafés and restaurants playbook. It fills the keyword field with what the trade is found under (cafes, coffee shops, restaurants, bistros, brunch spots) and weights the scoring so the report leads with AI visibility and structured data rather than with a page-speed number a cafe owner has no use for.
Sweep a single high street or neighbourhood at a time. Hospitality is intensely local, the competitor set on the report will be venues the owner can name, and that comparison does more work than any score.
The proof, in numbers
We publish what we have measured across every café and restaurant site this product has audited, category by category. It is a good page to send before a first conversation: it makes the point about the trade rather than about the venue, which is a much easier message to open with.