How restaurants get recommended by AI - and what to do about it now

For twenty years, the way a restaurant got found online was a list. Someone searched, got ten results, and chose from them. Ranking third was worth real money because third of ten still meant being seen.

Increasingly, people are not getting a list. They are asking an assistant where to eat tonight, near here, for this occasion, at this price - and getting two or three answers with reasons attached. There is no page two. Either you are one of the answers or you are not in the conversation at all.

That is a meaningfully different game to search, and it rewards different things.

Why this is not just SEO again

Search engines match a query to pages. Assistants answer a question, which means they need to know things about you rather than find documents mentioning you. The unit of visibility has moved from the page to the fact.

In practice, that changes what is worth having:

• Explicit, machine-readable facts about your venue - cuisine, price range, opening hours, neighbourhood, what occasions you suit - beat well-written prose about atmosphere.
• Consistency across sources matters more than volume. If your hours differ between your site, your Google listing and three directories, a model has no way to decide which is true, and the safest thing it can do is recommend someone else.
• Being described in the language people actually ask in - "good for a first date", "quiet enough to talk", "walk-ins welcome" - is what gets you matched to those questions. Nobody asks for "contemporary European small plates".
• Structured data on your own site (schema.org Restaurant markup, with hours, price range and menu) is the cheapest single thing you can do, and most independent restaurants still do not have it.

The five things worth doing this month

1. Fix your hours everywhere

   Your own site, Google Business Profile, and every directory that still lists you. Wrong hours are the fastest way to be filtered out of a real-time recommendation.

2. Add structured data to your site

   Restaurant schema with address, hours, price range, cuisine and a link to your menu. It is a one-off job for whoever built your site, and it makes your facts unambiguous.

3. Write the occasion, not the concept

   Somewhere on your site, say plainly who you are good for: date nights, big groups, solo lunch, before a show, kids. This is the vocabulary the questions are asked in.

4. Make the menu readable, not a PDF

   A menu locked in a PDF or an image is close to invisible to anything that reads the web. Plain text on a page.

5. Keep your reviews current

   Recency carries weight. A steady trickle of recent reviews is worth more than a large number of old ones.

What Planie is building here

Planie is a planning app: it builds someone's actual evening or trip, then decides which few places belong in it. That means a venue on Planie is not competing for a slot on a results page - it is being chosen for a specific person's specific moment, matched on occasion, vibe, time and distance.

Every listing is reviewed by a person before it goes live, and edits are re-checked, so the facts behind a recommendation are real ones. Partner and Featured tiers raise a venue's baseline visibility and campaign slots put it in front of a city at the right moment - but a boosted listing still has to clear the bar. Money can move you up; it cannot put you somewhere you do not fit. That rule is the reason being picked means anything.

Generative Engine Optimisation - structuring your venue so that assistants beyond Planie understand and recommend it - is the partner layer we are building next. Our placements page sets out how the whole thing works, and partner sign-up is open now.

It is also worth reading the other side of this: what AI gets right and wrong when it plans a trip is the same problem seen from the diner's chair.