Brands do not own their own facts. Whoever structures them does.

1. The observation

Three things in that paragraph matter more than the SEO joke. The query is a fact query, not a brand query. The winner has no kitchen and still gets the visit. And the visitor is let down at the end, because the winner cannot do the one thing the visitor came to do.

2. Why it happens

Brand websites are built to present the brand, not to answer questions about it. The menu is a PDF, an image, a JavaScript app, or an ordering flow behind a location picker. There is no page per question: no "breakfast menu", no "breakfast hours", no "menu prices". Search engines and answer engines need a page, or better a record, per question, with a date and a source. When the brand does not publish one, a third party does, with scraped and stale data, and it wins by default.

3. The vacuum is a market

Take one chain and list the questions people type after its name: menu, prices, breakfast menu, breakfast hours, nutrition, allergens, vegan options, kids menu, near me, delivery, coupons. Multiply by every chain, then by every city and every local restaurant, and the food vertical alone is millions of monthly fact queries that no brand answers on its own site. Answer engines make it worse for the brand and better for the record: a model that is asked "what is on the Tim Hortons breakfast menu" cites the page that is structured, fresh and attributable, not the prettiest one.

The same shape exists in every industry: bank charges, telecom plans, hospital packages, school fees, gym prices, courier rates. Menus are the wedge because the supply is the most public and the questions are the most frequent. Menuclature is the name for the method once it leaves food.

4. What "source of truth" means when the reader is a model

• Structured. schema.org Restaurant, Menu, MenuSection, MenuItem, Offer, and a plain JSON file with the same content.
• Dated and sourced. Every menu carries "as of", the URL it came from, the capture image and its hash.
• Verified. Extraction by the fleet, checked by an agent, spot-checked by a phone call or the restaurant's claim; the "menufied" mark means exactly that and nothing more.
• Queryable. menu.json, menu.md, llms.txt, a change feed and an API, so an agent never needs to parse our HTML.
• Honest about the order. The page's biggest button sends the visitor to the official site, app or phone. We take no orders. That is the fix to the tweet's complaint and the reason a brand will claim the page rather than fight it.

5. Why us

We already run the parts that are expensive to build for this idea alone. A corpus of four hundred million domain records with ownership, so we can enumerate every restaurant site and tell a chain's domain from a franchisee's. A capture archive going back years, so a menu has a history and a price has a previous value. A fleet that renders JavaScript and PDFs at scale. Agents that extract, normalise and question their own output. And the names: menufied.com for the record and menuclature.com for the layer beneath it. The Domain Fight Club can put Menufied in the ring in its directories month and let the scoreboard say whether the thesis holds.

6. The flywheel

Crawl the restaurant's own site. Extract and date the menu. Publish it AI-first on a page that answers the question. Rank and get cited. The restaurant notices the page, claims it and corrects it, which makes the record better than any scrape. Better records earn more citations, which bring more claims. Restaurants get a free, AI-ready page they could not build; we get the record. Claims are a bonus, never the supply: the primary rail must never wait for a human reply.

7. The moat

• History. Menus over time become a price index, "menuflation" by city and cuisine, which nobody scraping today can reconstruct.
• Coverage. The long tail through subsites, one per restaurant, generated from its own site.
• Trust. A mark that means dated, sourced and checked, and an order button that goes to the right place.
• Taxonomy. Menuclature: one identity for "masala dosa" across ten thousand menus, which is what makes comparison, nutrition and search work.

8. Risks, honestly

• Copyright. Items and prices are facts; descriptions can be expression. Keep descriptions short, paraphrased or quoted minimally, and give claimed brands the pen.
• Trademark. Nominative use with a clear "not affiliated" line on every page, a claim flow, and an opt-out that is honoured within a day.
• Search engines' view of scraped content. Add what the source lacks: structure, dates, history, comparison, verification, the official order path. A page that is only a copy deserves to lose.
• Staleness. Recrawl cadence by volatility, "as of" on every page, claims to override, and a public change feed so being wrong is visible and short-lived.
• Platforms. Zomato and Swiggy hold menus for ordering; we do not compete on ordering. We are the record, and we link to them as an order path when the restaurant is there.

9. The rules we keep from offer.in

Never touch money. Never a percentage of anyone's sales. No sales army: restaurants claim self-serve. The index complete and free; ranking and curation only in the human view. Agent-first: JSON-LD, llms.txt, feeds, an MCP endpoint. And the supply rail reads the restaurant's own published data, so it never waits for a founder to reply.