How Restaurants Show Up When Someone Asks AI for a Recommendation

The query has changed shape

Think about how someone actually asks for a restaurant recommendation now. A few years ago it was ‘Italian restaurants Manchester’ typed into Google, followed by a scroll through a map pack of ten pins. Now it’s increasingly a direct question typed into ChatGPT or spoken to a phone: ‘where’s a good quiet Italian for a date night near Deansgate that does a proper carbonara’. Nearly half of all Google searches now surface an AI answer directly on the results page, 48% as of this year, before a single blue link even appears. That’s not a keyword, it’s a conversation, and the AI has to actually answer it, not just list options. It’ll often name one or two places outright rather than presenting ten. If you’re not one of the one or two, you don’t get a slot on page one where a determined browser might still scroll down and find you, you just don’t get mentioned. This matters more in hospitality than almost anywhere else because eating out is such a spontaneous, trust-led decision. Nobody weighs up fifteen options for a Tuesday night curry in Rusholme. They ask, they get an answer, they book. Being the answer, not the ninth blue link, is now the whole job.

What the AI is actually checking

So what’s the AI actually checking before it puts your name in its answer? Mostly, consistency and specificity. It’s cross-referencing your opening hours, menu, price range, and location across every platform it can see, your own website, Google Business Profile, TripAdvisor, Just Eat or Deliveroo listings, even old directory sites you forgot existed. If your website says you close at 10pm but Google says 9pm and TripAdvisor says nothing at all, that’s a small trust wobble that adds up. It’s also reading review sentiment, not just your star average. A 4.2 rating with reviews that specifically mention ‘friendly staff’, ‘quick service on a Friday night’, or ‘brilliant with kids’ gives the AI concrete detail to work with, far more useful than a generic 4.6 with one-word reviews. Freshness matters too. A restaurant with reviews from three years ago and nothing since looks, to a model checking for current, reliable information, like it might have closed. Recent reviews, even a handful a month, tell the AI you’re open, active, and worth recommending today.

Why one glowing review site isn’t enough

Here’s the trap a lot of good restaurants fall into: they’ve got a brilliant TripAdvisor page and assume that’s job done. It isn’t, because AI systems triangulate across sources rather than trusting any single one. If your Google Business Profile is half-filled-in, your website hasn’t been touched since before the pandemic, and your Instagram bio has an old address on it, that inconsistency undermines the strong TripAdvisor page rather than sitting quietly alongside it. Think of it less like one report card and more like several witnesses being asked the same question. If they all say the same thing, confidently, the AI trusts the answer. If they contradict each other, even slightly, the model tends to hedge or leave you out rather than risk sending someone to a restaurant that might have moved or shut. For a place in somewhere like the Northern Quarter or Chorlton with three or four listings that haven’t been touched in years, that’s often the single biggest fix available, not new marketing, just making the existing information agree with itself everywhere it appears.

What to actually do this month

If you run a restaurant, café, or bar and want to actually move the needle this month, start boring and specific. Update your Google Business Profile properly: correct hours including bank holidays, an accurate menu with prices, decent photos taken this year not five years ago. Then check that every other listing, TripAdvisor, Facebook, delivery apps, your own site, matches it exactly. Ask happy regulars for reviews that mention something specific, the dish, the occasion, the thing that made it good, rather than just ‘great food, 5 stars’. Reply to reviews, good and bad, because that activity signals a business that’s actually open and paying attention. And if your menu changes seasonally, make sure the online version changes with it rather than sitting stale for months, since an AI reading last spring’s menu in November is going to give a customer the wrong expectations, and a wrong expectation is worse for you than no answer at all.