Restaurant SEO in Nice: the Local Pack, your Google profile, your menu and AI answers to "where to eat"

Restaurant SEO in Nice is won on Google Maps first: a complete profile, a menu written as text and recent reviews decide whether you make the Local Pack, the three places shown on the map. The same information then feeds the answers ChatGPT, Gemini or Perplexity give when a visitor asks where to eat in Old Nice.

For a restaurant the job is fairly small: help machines understand quickly what you serve, where, when and at what price, then get other sources to confirm it.

How do people pick a restaurant in Nice today?

They search on their phone, often a few hundred metres from your door. “Socca restaurant Old Nice”, “brunch Nice port”, “restaurant open Sunday evening Nice”: these searches show a map before any link to a website.

Google’s help page on local ranking names three factors: relevance (does your profile match the search), distance, and prominence, meaning how well known you are, measured among other things by links to your site and the number of reviews. Google also states that you cannot pay to rank higher.

More and more of the journey goes through AI. BrightLocal’s Local Consumer Review Survey 2026, run with 1,002 US consumers and published in February 2026, found that 45% had used ChatGPT or another AI tool for local recommendations, up from 6% a year earlier. These are US figures, but the visitors landing in Nice from New York or Toronto bring those habits with them.

Which Google profile settings matter for a restaurant?

A restaurant’s Google Business Profile has options other trades don’t get. These are the ones I set first.

Setting	What Google says	What I recommend in Nice
Primary category	it shapes which searches you appear for	the most specific one Google offers for your food (“Pizza restaurant”, “Seafood restaurant”) rather than “Restaurant”
Menu editor	for “Food and drink” businesses, with dish name, description and price	type the dishes as text, by section, using the names people search for (socca, pissaladière, daube)
Menu link	added in the contact information	point to an HTML page on your site, not a PDF
Reservations	through a partner provider or a custom link	use the same tool everywhere (profile, site, social)
Special hours	public holidays, events	Carnival, the end-of-year holidays, annual closure
Photos	customers add some too	dining room, terrace, real dishes, the frontage as seen from the street

The menu editor deserves a closer look. According to Google’s help, changes can take 24 to 48 hours to show, Google can also transcribe the menu automatically from your website, and an experimental feature turns a photo or PDF into a detailed menu. Always check the result: a wrong price on Google turns into a bad surprise at the table, then a bad review.

Categories, address and reviews are covered in my Google Business Profile guide for Nice, with a 15-point checklist.

Why publish your menu as text on your website?

A menu as a photo or a scanned PDF is hard to read for Google and for AI tools alike. An HTML “Menu” page with dishes, prices and useful notes (vegetarian, gluten free, dish of the day) answers precise searches directly.

A few simple rules:

1. One stable menu URL, updated in place rather than replaced every season.
2. Dish names the way customers type them. If you serve petits farcis, write “Niçoise stuffed vegetables (petits farcis)”, not just “seasonal vegetables”.
3. Current prices, identical on the site, the Google profile and booking platforms.
4. An English version at its own URL if you serve tourists. On the Cours Saleya or around the port, that’s a large share of the room in summer.

The site itself can stay small. A home page that states the cuisine and the neighbourhood, a menu page, a booking and contact page, an English version: for most restaurants that’s enough.

Which structured data should a restaurant add?

Structured data is a block of invisible code (JSON-LD) that describes the business to search engines. For a restaurant you use the Schema.org Restaurant type. Google’s local business structured data documentation requires name and address, and recommends among others servesCuisine, menu (the URL of your menu page), openingHoursSpecification, priceRange, telephone and geo.

Schema.org also has acceptsReservations, which Google doesn’t list but other engines can read. What matters is that the markup says exactly what the visible page says. A full annotated example is in the LocalBusiness schema markup guide.

How do you get reviews without breaking Google’s rules?

Ask every customer, at the right moment, and offer nothing in return. Google’s contribution policy bans any benefit in exchange for a review (a discount, a free dessert, a free drink) and selectively asking only happy customers. Reviews obtained that way get removed.

What works in the dining room:

• a QR code on the bill or bill folder that opens the review form directly;
• a spoken request from the server at the end of a meal that went well, without picking which tables to ask;
• a reply to every review, bad ones included, with concrete details rather than a pasted template.

Freshness counts as much as volume. In the same BrightLocal survey, 74% of consumers look for reviews written in the last three months. A restaurant with 900 reviews, the latest from last winter, inspires less confidence than a neighbour who gets new ones every week.

Reviews that describe the context help AI tools too: “quiet terrace in Old Nice”, “proper socca, nicely charred”, “staff spoke English”. You can’t dictate those words, but you can ask guests to say what they enjoyed.

How do ChatGPT, Gemini and Perplexity answer “where to eat in Nice”?

These assistants build an answer from sources they find online and keep the names that come up in several of them.

When ChatGPT searches the web, it rewrites the question into several queries and, according to OpenAI’s help, can rely on third-party providers for local information such as business listings. On Google’s side, Ask Maps, the Gemini-based conversational feature in Google Maps, launched in March 2026 in the US and India, as reported by TechCrunch. As of this article I haven’t found an official Google announcement for France.

Here are the kinds of questions to test yourself, in English and in French:

Question asked to the AI	What it tries to check
”Where can I eat real Niçoise food in Old Nice?“	cuisine type, neighbourhood, sources that talk about tradition
”Restaurant open Monday evening near Nice port”	exact hours, consistent everywhere
”Good seafood restaurant in Nice with a terrace”	English content, reviews that mention the terrace
”Restaurant for a group of 15 in Cimiez”	capacity, private room, group bookings
”Vegetarian brunch in the Libération area”	menu as text, dietary notes

To weigh in these answers you need to exist in third-party sources: local press, guides, travel site round-ups, review and booking platforms. In Nice, the Cuisine Nissarde label is one of them. It’s run by the Nice Côte d’Azur Tourist Office, which publishes the list of restaurants recognised for traditional recipes and local produce. If you cook genuine Niçoise food, that kind of recognition is worth more than a dozen directories.

The general method is in the article on local GEO, and how to track results over time in the one on measuring AI visibility.

Which neighbourhoods and searches should you target in Nice?

People search by neighbourhood and by craving. Your site and profile should name the area the way locals do, and describe what people find at your place.

• Old Nice and the Cours Saleya: heavy tourist demand, searches in English and Italian, queries about socca and Niçoise food.
• Port Lympia: seafood, aperitifs, terraces in the evening sun.
• Libération and Borriglione: local regulars, lunch, the market, brunch.
• Carré d’Or and the Promenade des Anglais: hotel guests, dressier dinners, online bookings.
• Cimiez and Mont Boron: family meals, groups, parking.

A page per neighbourhood only makes sense if you have several locations. For a single restaurant, say clearly where you are and what’s nearby (“two minutes from the Cours Saleya market”).

Where should you start?

If I had one week for a restaurant in Nice, I’d go in this order:

1. Google profile: a specific category, the menu typed as text, menu link, reservations, special hours, 20 recent photos.
2. Website: an HTML menu page, an English version, the same name, address and phone number everywhere.
3. Restaurant structured data that matches the page.
4. A way to ask for reviews in the dining room, and a reply to every review.
5. A list of the sources that already cite your competitors in ChatGPT and Perplexity, so you know where to look for mentions.

I do this work from Nice, and remotely for businesses across France and abroad. The Google + AI visibility audit (€390, delivered in 5 days) shows where you stand on Maps and in AI answers. Monthly SEO/GEO support (€490 a month) keeps the profile, reviews and site up to date all year. Everything is on the SEO & GEO page, and the overall approach is in the article on local SEO in Nice.

Frequently asked questions

Does a restaurant need a website if its Google profile is complete?

Yes, a simple site is still useful. It holds the text menu page, the English version and the structured data, which Google and AI tools read to confirm what the profile says. Four or five pages are enough for most restaurants.

Can I offer a free dessert in exchange for a Google review?

No. Google bans any benefit, direct or in kind, in exchange for a review, and removes reviews obtained that way. You can ask every customer for a review, with nothing in return and without filtering tables.

How long does it take to climb the Local Pack in Nice?

Profile changes show within 24 to 48 hours, but the effect on rankings depends on competition in your area and how steadily reviews come in. Think in months rather than weeks in Old Nice, where competition is very strong.

How do I know whether ChatGPT recommends my restaurant?

Ask it your customers’ questions, in English and French, several times a few days apart, and note who comes up and which sources are cited. A panel of 10 to 20 questions tracked monthly gives a reliable trend.