How restaurants actually get found on Google in 2026

I spent sixteen years running multi unit restaurants before I did this work, and the thing nobody told me then is that the search result is an operational asset, not a marketing one. It has hours in it. It has photos of your dining room. It has whatever your last twelve guests said about a Tuesday.

That means the search result drifts the same way your operation drifts. Nobody breaks it on purpose. It just stops being true.

What is Google actually using to decide which restaurants to show?

Three things, and only one of them lives on your website.

Where the searcher is standing. Proximity of your address to the point of search scored second out of 187 factors in Whitespark's 2026 Local Search Ranking Factors survey, in which 47 local search experts rated every input they could name. You cannot buy your way out of it. A restaurant four blocks closer to the person holding the phone starts ahead of you.

Your Business Profile data. Category, hours, attributes, menu, service options, photos. The same survey puts primary category at number one, keywords in the business title at three, a physical address in the city of search at four, and whether you are open at the moment of search at five. Four of the top five are profile fields. Most restaurants filled them in once during opening week.

Your reviews, especially the recent ones. Not just the star average. Recency of reviews ranks eleventh in the survey and a sustained influx rather than bursts ranks fourteenth. A profile that stopped collecting reviews in March reads as a business that may have stopped operating.

Grade that survey properly. Whitespark sells local SEO software, and the report is expert opinion, not controlled testing. Darren Shaw, who runs it, states plainly that the panel has no access to Google's algorithm. The top four have held their order across editions anyway.

When a place with worse food sits above you, it is almost always one of those factors rather than a judgment about the cooking, which is why worse restaurants outrank better ones so often.

Why does my website matter less than I expected?

Because the two results answer different questions.

Someone searching "korean bbq near me" at seven on a Friday gets a map. Three listings, photos, hours, a call button. Most never scroll to the blue links underneath. The decision happens inside the map card.

Someone searching your restaurant by name is already sold and wants a phone number or a reservation. That search should land on your site, and often it lands somewhere else instead. If a third party is winning your own name, that is fixable, and it is the same problem as a delivery app ranking above your own website.

But your site is not outside the local ranking system either. The same Whitespark survey scores keywords in the profile's landing page title tag at number 17 and keywords in that page's headings at number 20. The page your profile points at is read as part of the profile. A thin homepage weakens the asset above it.

So build the site for the person who already found you and now needs a fact.

Which searches are actually worth showing up for?

Not the ones you would guess. Cuisine plus neighborhood is obvious and crowded. The searches that convert best for restaurants tend to be narrower.

Occasion searches. Private dining, large party, birthday dinner, group of twelve. The person searching has a date and a headcount and will book something this week.

Constraint searches. Gluten free, halal, open late, patio, parking. These read like small niches and they are the searches where a specific answer wins outright, because most restaurants never publish the answer.

Catering and events. A different buyer entirely, with a much larger ticket, and a buyer who evaluates you in a completely different order than a diner does.

The pattern is that you win where you can answer plainly and your competitors publish nothing.

Does any of this change now that AI answers exist?

Less than the category wants you to believe, and the numbers are checkable.

Google's own documentation states it directly: "From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO." Google also says your Business Profile helps a local business appear "in both AI responses and other Google Search results." The same record feeds both.

Now the volume. Ahrefs analyzed 146,122,391 desktop SERPs and found AI Overviews on 20.5% of all searches, but on only 7.9% of local searches. Sterling Sky, tracking a different keyword set with different tooling, independently observes AI-powered local packs on about 7% of tracked keywords. Two methods, two vendors, the same ballpark.

There is a real counterpoint and it should travel with those numbers. A Whitespark study of 540 manually run queries across six local industries reported AI Overviews on roughly 68% of local business queries. Both are right about their own sample. Ahrefs classified local by intent across a database dominated by short "[service] [city]" patterns that rarely trigger one. Whitespark hand-picked long conversational queries that almost always do. Short transactional local searches are still local pack territory. Long "help me choose" searches are AI territory.

Separately, SparkToro and Similarweb's US panel found only 0.34% of Google searches transitioned into AI Mode between January and April 2026.

What does change is the answer format. An AI response summarizes rather than listing ten options, so the facts on your profile and your site need to be unambiguous. If your hours say one thing and your site says another, you have handed a summarizer a coin flip.

Can an agency rank me first in ChatGPT?

No, because there is no rank to be first in, and this is the clearest place in restaurant marketing where the sales pitch and the evidence separate.

Researchers at the University of St. Gallen ran identical prompts across ChatGPT, Gemini, Google AI Mode and Perplexity every day for 45 days, and then ran the same prompt ten times in a single day. Between two consecutive days, only 34% to 42% of cited sources overlapped. Same day, same prompt, repeated runs: source overlap of 32% to 43%. ChatGPT was the least stable engine measured. It also activated web search on only about 42.2% of runs, meaning 57.8% returned no citations at all.

There is no position. There is a probability, and it changes overnight.

The paper's prescriptions are usable as a buying test. Seven or more runs per prompt per day for a brand visibility figure. A rolling two to four week window, never week over week. A large, diverse prompt set, because per-prompt variation swamps everything else. If a tool cannot tell you its number of runs, its confidence interval and its prompt list, it is selling you noise with a chart on it.

Google's own warning is blunter than anything I would write: "Be wary of third-party tools that promise ranking success or claim to use 'internal' Google metrics. No third-party tool has access to our internal ranking or AI systems."

What is the AI channel actually worth to a restaurant right now?

Here is the arithmetic. Use your own analytics numbers in place of mine.

Sterling Sky published the cleanest available real-world ratio. For one of its largest multi-location clients, traffic from ChatGPT went from 0.1% of the client's Google traffic in 2025 to 2% in 2026. Large growth rate. Small absolute number. And it was still only 22% of what Bing sent that same client.

Apply that shape to a restaurant site doing 4,000 sessions a month from Google:

• ChatGPT at 2% of Google traffic is about 80 sessions a month.
• Bing, at the ratio above, would be roughly 360 sessions a month.
• A $1,500 monthly retainer sold as AI visibility work is therefore priced at about $19 per session against the channel it claims to influence, before anyone establishes that the work moved the number.

Run it with your own Google session count. If the answer is single digits of sessions, you have your decision.

Two qualifications. That ratio is one agency's client, not a benchmark, and a restaurant's mix will differ from a multi-location professional services brand. And the direction of travel is real: SparkToro's panel put 68.01% of US Google searches ending without a click in 2026, up from 60.45% in 2024. Zero-click is growing. It is not yet growing into a channel you can buy your way into.

Claim your Bing Places listing anyway. Free, twenty minutes, and every credible account of ChatGPT's local index runs through Bing.

What is actually changing in local search this year?

Four things, all documented, none of them AI.

Sterling Sky's field data shows local pack ads went from about 1% of tracked mobile reports in early 2025 to about 22% in December 2025. Local Services Ads went from about 11% to 31% of tracked queries. Joy Hawkins's summary: "Google is definitely going pay-to-play."

Clicks-to-call from Business Profiles are declining even for profiles whose rankings held. Jepto compiled 179 profiles across 34 US law firms over two years to show it. Website clicks did not fall the same way, because the call button dominates mobile and the website icon dominates desktop. The loss is mobile-specific and it is the call button, which for a restaurant is the conversion.

And where AI local packs do appear, they show one or two businesses instead of three and carry no call buttons. Places Scout counted 5,943 unique businesses across AI local packs versus 18,330 in regular three-packs, about 32% as many. In 322 markets, 88% had fewer unique businesses in the AI version.

That is the real 2026 story. The shelf is getting narrower and more of it is for sale.

What to do this week

Open your Google Business Profile and read it as a stranger would. Hours including holidays, primary category, service options, phone, menu link, and the last date a photo was added.

Then search your own restaurant name on a phone that is not yours, from somewhere other than the restaurant. Note what comes up first. If it is a delivery marketplace or an aggregator, write that down.

Then check your hours against your actual schedule for the next month. Vacations, holiday closures, the Monday you stopped opening for lunch. Wrong hours cost you more than any ranking change, because a customer who drives to a closed restaurant leaves a review about it.

Then claim Bing Places. Twenty minutes, no downside.

If you rank fine most of the day but vanish during service, that is a specific and common failure with a specific cause, covered in why a restaurant disappears from Maps at dinner time.

Questions about how your restaurant shows up?

Email me at eric@seod.com with your restaurant name and city. I will run the searches your customers actually run, from a phone in your neighborhood, and send you the screenshots of where you land. Not a report, just the pictures and two or three sentences on what I would fix first.

I ran restaurants for sixteen years and I answer these myself. If the honest read is that your profile is fine and your problem is somewhere in the operation, that is what you will get back.

Or keep reading more on restaurant marketing.