The classic layer: the local trinity
Three assets decide restaurant demand and none is a secret: an immaculate business profile, menu, hours, photos, attributes, kept current because wrong hours are the fastest one-star generator in the industry; review velocity with responses, since diners read the newest five reviews and your replies before anything else; and direct-order economics, shifting even a fraction of delivery volume from marketplace commissions to first-party ordering pays for the whole marketing program. Email and SMS to your own list beat every paid channel for repeat visits, and repeat visits are the entire economics of the business.
What burns cash: broad-radius paid social and discount-app dependency that trains customers to never pay full price.
The 2026 layer: the assistant picks the shortlist
Where should we eat tonight has become an assistant question: date night spot with good vegetarian options near the theater, best ramen open late, somewhere quiet for a work lunch that takes reservations. ChatGPT, Gemini, Claude, and Perplexity answer with three to five named restaurants, synthesized from reviews, local coverage, menus, and community threads, the same shortlist dynamic from why it is so hard to show up in AI, played nightly in every neighborhood.
The restaurant inputs are refreshingly concrete: a real menu page in text, not a PDF image, with prices, because assistants cannot recommend what they cannot read; attributes stated plainly, dietary options, noise level, reservations, patio, kid-friendliness, since those match the situations diners describe; reviews that mention dishes and occasions; and presence in the local threads where your city argues about food, per Reddit's role in AI answers.
A 90 day plan for an owner-operator
Days one to fifteen: baseline and hygiene. Ask the assistants the questions your actual diners would, your cuisine plus your neighborhood plus real situations, log who gets named, run the free GEO Grader, and fix the profile layer: hours, menu, photos, attributes. Days fifteen to fifty: publish the readable menu with prices and a short page per occasion you genuinely serve well, groups, dates, family dinners, work lunches; start the review ask at the table or on the receipt. Days fifty to ninety: earn one local write-up or listicle inclusion, engage honestly in the neighborhood threads, re-baseline. Restaurant demand cycles weekly, so this is the rare category where visibility movement shows up in covers within the quarter.
Measuring covers, not likes
Ignore follower counts; track what fills seats: assistant share of voice for your occasion-and-neighborhood questions monthly via AI search tracking; direct orders and reservations as a share of total, the margin metric; how-did-you-hear at booking, with an AI option, which staff can log in one tap; and repeat rate from your own list. A restaurant that owns its neighborhood's assistant shortlist and its own ordering funnel has replaced most of what it used to rent from platforms, which is the whole strategic point.
