Restaurants Are Building AI Websites — and Answer Engines Are Reading Them
Of the 67 websites blasa has published so far, 48 are restaurants, cafés or bars. What changed this year is not that they get a website — it is who reads it: alongside Googlebot, the crawlers behind ChatGPT, Claude, Perplexity and Gemini now request those pages thousands of times a fortnight.
Restaurants are most of what gets built
As of 20 August 2026, blasa has 67 live business websites. 48 of them are restaurants, cafés or bars — roughly seven in ten. They are spread across countries rather than clustered in one: La Mega Pizza in Chelghoum Laïd, عرفة اكسبرس in Irbid, Panorama Cafe in Faraya, Falaknama in Palia Kalan. The full list is public in the business directory.
There is a reason food businesses dominate. A restaurant's reviews are unusually rich raw material: people name the dish, the time of day, who they came with, what they would order next time. That is exactly the kind of specific, quotable detail a generated page can be built from — and exactly what a scanned menu never contains.
Who is actually reading these pages
The origin logs every bot that identifies itself. Over a rolling 14-day window ending 20 August 2026, crawlers made 26,418 requests to blasa and its hosted sites. 11,786 of those came from AI crawlers — not search engines, but the agents that feed answer engines.
The named ones, by request count over those 14 days:
- ClaudeBot — 1,509
- GPTBot (ChatGPT) — 1,462
- OAI-SearchBot (ChatGPT search) — 1,321
- PerplexityBot — 334
- Google-Extended (Gemini / AI grounding) — 226
Classic search is still there and still matters: Googlebot made 1,308 requests and Bingbot 691 in the same window. The honest caveat: being crawled is not the same as being cited, and a user-agent string can be forged. But nothing gets cited that was never crawled — so this is the first gate, and it is one you can verify for your own site instead of guessing.
Why a review-built page is easy for an answer engine to quote
Answer engines are lazy readers, and that is the whole design constraint. Four things decide whether your page is usable to them:
- It renders without JavaScript. Every blasa site is prerendered static HTML — the text is in the markup, not assembled in a browser.
- It says what kind of business it is, in machine-readable form. LocalBusiness structured data with address, hours, geo and rating, plus FAQPage markup for the questions.
- It is specific. Dishes customers actually name, in the words they used. "Great food" is unquotable; "the mixed grill for two on a Friday" is not.
- It publishes an
llms.txt. A plain-text summary of the business and its FAQ, at a fixed path, for agents that prefer to read that first.
What a restaurant owner actually has to do
Paste your Google Maps link. The site is generated from your reviews, photos, logo and hours, and published on its own address within minutes — nothing to write, no template to choose, no editor to learn. From there the useful work is the same work you were already doing: keep earning reviews that name things. Every new review that mentions a dish, a neighbourhood or an occasion is more material the page — and anything reading it — can use.
If you want the longer version of how the page itself is assembled, read what a restaurant gets from an AI website built on reviews or how reviews become a business website.
Questions people actually ask
Does ChatGPT read my restaurant website?
It can, through two crawlers: GPTBot (used to build the model's knowledge) and OAI-SearchBot (used to fetch pages for search answers). Over the last 14 days those two requested pages from blasa-hosted sites 1,462 and 1,321 times respectively. Neither can read a page it is blocked from, and neither runs JavaScript reliably — which is why the pages have to be static HTML.
How do I get my restaurant mentioned by Gemini or Google AI Overviews?
Google grounds those answers in pages it has crawled, and it uses a separate agent — Google-Extended — for AI training and grounding access. There is no way to pay for a mention. What you can control is being crawlable, being specific (dish names, neighbourhood, hours as real text), and having structured data that says plainly what kind of business you are and where.
I already have a Google Maps listing. Do I still need a website?
A Maps listing answers "is this place open and where is it". A website is the page an answer engine can quote sentences from — your dishes, your story, your FAQ — and the only one of the two you actually own. Restaurants that have both get named in more of the answers, because there is more to name.
What language should a restaurant website be in?
The one your reviews are in, plus English. blasa detects the dominant language of your Google reviews and builds the site in that language with English alongside — right-to-left layout included when the reviews are in Arabic. Among the live sites today the review languages include English, French, Arabic, Spanish, Portuguese and German.