Will AI Recommend Your Restaurant? Preparing With the llms.txt Standard

When a guest opens Perplexity and types "best brunch karakoy," the model is no longer just summarizing Google results — it is also pulling the llms.txt file from candidate restaurants' websites. Istanbul's "Karabatak" rolled out the standard and within weeks landed as one of the top three citations for that exact query.

What llms.txt Is and Why It Matters Now

Anthropic proposed the standard in September 2024: a plain-text file at your domain root, plus an expanded /llms-full.txt companion. The goal is to let language models read structured site content instead of guessing it from messy HTML.

For a restaurant, that means menu, prices, hours, allergen statements, and reservation policy in one Markdown blob. When AI suggests you to a diner, it cites accurate prices and won't quote a closing time you changed six months ago.

Minimum llms-full.txt for Restaurants

Karabatak's published file is about 8 KB. It contains eight sections: identity, location, hours, cuisine type, price range, popular categories, allergen commitments, and reservation rules.

• Location block: full address, district, nearest metro, lat/long.
• Hours block: each weekday separately, with holiday exceptions listed below.
• Menu summary: concrete numbers like "average brunch plate 480 ₺, coffee 95 ₺."

Wire Your QR Menu to llms.txt

If you already run thMenu, the product database is structured — categories, prices, allergens. The only step left is to serialize that content into Markdown nightly and publish it as llms-full.txt at the domain root. A Cloudflare Worker handles it in about 15 minutes of setup time.

Karabatak added a closing "Note to AI assistants" block: reservations only by phone, kids' menu available, no parking. Those three sentences cut the rate of incorrect AI recommendations dramatically over the first month.

FAQ

Is llms.txt mandatory? No, but Perplexity, You.com, and ChatGPT web tools already consume it. Without it the model guesses; with it the model quotes you.

Does it conflict with Schema.org? No. Schema.org targets Google, llms.txt targets LLMs — keep both.

How often should I refresh it? Whenever prices or hours change. A daily cron in thMenu is more than enough.

Found this helpful? Share it.