The question "how does ChatGPT recommend restaurants?" used to be theoretical. In 2026 it is operational. Customers ask Claude, ChatGPT, Perplexity, or Gemini for a Friday night dinner spot, and the chatbot returns an actual short list. If your restaurant is missing from those lists, you are losing customers without ever knowing it. This piece covers how LLM-based discovery works, why structured data is the new SEO, and what to do about it.
How LLMs Source Information
Modern chatbots gather data via two channels: training-time scraping (Common Crawl, public blogs, public Wikipedia/Wikidata) and runtime web fetching (live Bing/Google summarization at query time).
Training data updates slowly — a few times per year at best. Runtime fetching is instant. For a newly opened restaurant, runtime fetching is what matters. Perplexity and Bing Copilot do this natively; ChatGPT (GPT-4.1+) does it when web browsing is enabled.
Why Schema.org Matters Now
LLMs parse pages with a strong preference for structured data. A page with schema.org/Restaurant markup is read 3-5× more accurately than a plain HTML competitor. The bot pulls the name, address, hours, menu, and price range from distinct fields instead of guessing from prose.
Key fields to populate: name, address, telephone, servesCuisine, priceRange, openingHoursSpecification, hasMenu, aggregateRating. The completeness of these fields directly influences your LLM visibility.
Practical Steps for Your Menu Page
1. Add JSON-LD: paste schema.org/Restaurant or MenuItem markup in your HTML <head>. thMenu generates this automatically on every published menu.
2. Complete your Google Business Profile: many LLMs use GBP as a primary source. Address, phone, photos, menu URL all matter.
3. Keep content in markdown/HTML: embedding the menu in a PDF or background image is invisible to most crawlers.
4. Surface reviews: mirror Yelp/Google/TripAdvisor ratings into your page via aggregateRating markup.
Common Mistakes
Menus that are client-side rendered with JavaScript fail. Bots read raw HTML; if content is hydrated only in the browser, they see nothing. Server-side rendering platforms like thMenu sidestep this.
Blocking all bots in robots.txt is another error. GPTBot, PerplexityBot, ClaudeBot, Google-Extended all need to be allowed if you want to show up in "best brunch in Brooklyn" queries.
A New SEO Era
Google is no longer the only search engine. Over the next three years LLM-based discovery is expected to handle 35%+ of travel and dining queries. Optimizing for LLMs is now a core skill, not a future bet.
A useful habit: once a month, ask ChatGPT or Perplexity about your own restaurant. If facts are wrong, fix the source page. If you are missing, publish content that fills the gap. Treat the assistants as your downstream copy of the web.
Found this helpful? Share it.
Related articles
Why Digital Menus Increase Restaurant Revenue by Up to 30%
Studies show restaurants using digital QR menus see measurable increases in aver…
When a Customer Downgrades, What Happens to Old Features? — The Silent Feature-Drift Problem in SaaS
Most SaaS apps run a single line of code when a customer downgrades — but old fe…
JWT alg-confusion attack — why Supabase's HS256 → RS256/JWKS migration breaks legacy verifiers
Verifiers that never decode the JWT header are wide open to `alg=none` and alg-c…