Multilingual QR Menu for Restaurants: AI Translation and UX That Works

Your guests are reading the menu.
But they don't understand it.

793 million tourists visited Europe in 2025. Most of them faced a menu in one language — and chose a familiar dish instead of what the establishment wanted to offer.

Sunday evening. A tourist restaurant in the centre of Florence. At one table — a couple from Poland, next to them — a family from Germany, by the window — two guests from the Netherlands. On each table — one menu. In Italian.

The waiter waits. Guests open their smartphones. Google Translate slowly processes the photo. Someone taps at random on an unfamiliar name. Someone asks for 'just anything' because it's awkward to ask again. The average bill is modest — not because guests didn't want to order more, but because they didn't understand what was on offer.

This is not an isolated situation. This happens every evening in thousands of restaurants across Europe. And it is fixable.

Why this matters more than you think

Tourists are not a random audience for a European restaurant. According to the World Food Travel Association, 25 to 35% of the entire tourist budget goes on food and beverages. This is not a secondary expense item: it is the main one. Dining is one of the key reasons people travel at all.

In 2024, the EU recorded 2.99 billion tourist overnight stays — a historic record according to Eurostat. Foreign guests alone accounted for 1.43 billion nights, 58.7 million more than the previous year. Italy and Spain remain top destinations: the three largest markets account for 46% of all nights spent abroad.

Now the main question: if a guest spends five nights in your city, spends a third of their budget on food, but doesn't understand the menu — what share of that budget do you receive?

What goes wrong — and why familiar solutions don't work

Problem 1. A printed menu in five languages is not a solution

The first instinct of many restaurateurs is to print a menu in several languages. The logic is understandable: if a guest reads German, give them a German menu. The problem is that this is not solving the problem, but scaling it.

Every price change, every seasonal item, every new dish is multiplied by the number of languages and print runs. According to WAND Corporation, a typical restaurant with monthly updates spends $200 to $500 on a single print run at one location. With four seasonal menu changes per year, that's $800–2,000 per language. Multiply by five — and you get a sum that already requires a separate budget line.

But money is not the main problem. The main one is speed. A dish has run out? In a digital menu that's fixed in seconds. In a printed one — it waits for the next print run. Every delay — it's the waiters forced to explain discrepancies between the menu and reality — with a language barrier on top.

Problem 2. Google Translate turns your menu into a comedy

When a guest photographs a menu and runs it through an automatic translator, something quiet and destructive happens: they receive someone else's version of your restaurant. Not the one you've been creating for years.

'Guanciale' becomes 'pillow'. 'Contorni' becomes 'contours'. A dish loses its story before it's even brought to the table.

This is not an exaggeration or an isolated curiosity. Automatic translation without culinary context systematically fails on several types of content: regional names without a direct equivalent, non-literal historical terms, chef's signature concepts, abbreviations and menu shorthand. The result — the guest orders something unexpected, is disappointed, and writes about it in a review. Not about the food — about the misunderstanding.

Problem 3. The language barrier works both ways

When a guest doesn't understand the menu, they don't just order randomly. They order less. The logic is simple: if it's unclear what 'antipasto della casa' is, it's safer to take one familiar dish and not risk the unfamiliar ones. Fewer clarifying questions are asked — because it's awkward to ask when the waiter doesn't speak your language. Desserts and drinks are skipped — because by that point they're already tired of guessing.

CSA Research (8,709 respondents, 29 countries) shows: 76% of consumers prefer information in their native language, and 40% won't make a purchase at all if it's unavailable in their language. A menu is exactly such a decision point.

AI translation vs manual: what actually works

When it comes to translating a menu, restaurateurs usually face a false dilemma: either an expensive professional translator for each language, or Google Translate with its comedic results. In reality, the optimal solution lies between the two.

Where AI does well

Modern language models work excellently with most menu items: standard cooking techniques (grilled, baked, marinated), common ingredients, descriptive adjectives, structural elements (starters, mains, desserts). For a typical menu of 40–60 items, AI produces a quality draft in any European language in minutes, not days.

Where AI goes wrong — and why it matters

Problems start in several predictable places. Regional dishes with historical names — 'brandade', 'bistecca alla fiorentina', 'bouillabaisse' — are translated literally or transliterated. Chef's signature dishes containing wordplay or a cultural reference lose their meaning. Dialect terms and local abbreviations give unpredictable results.

Hybrid approach: one hour of work instead of a week

The optimal scheme: AI generates a draft of the entire menu, the chef or a native speaker checks 10–15 items with non-literal names. For most restaurants this takes an hour. Updates when a seasonal dish changes — minutes, without involving a translator.

When a professional translator is still needed: fine dining menus with extended descriptions that are part of the restaurant's concept; legal texts (allergens, ingredients, warnings). Everything else — the hybrid approach covers.

How many languages you need — and how to find out without guessing

This is one of the most frequent questions, and the answer is usually not what's expected. Most restaurateurs either underestimate ('English is enough') or overestimate ('we'll take twenty at once'). Both approaches are imprecise.

Method: look at data, not intuition

When a guest scans the QR code, the platform automatically detects the language of their phone. This is not a perfect metric (a tourist with a German phone may be French), but accurate enough for decision-making.

This also means starting with three to four languages is reasonable. You won't guess the ideal set in advance. But you'll quickly refine it based on data.

Starting priorities for European destinations

Before your own data appears, Eurostat statistics provide guidance: in 2024, Italian and Spanish restaurants hosted tourists primarily from Germany, France, the Netherlands and Poland. Balkan destinations — Croatia, Greece — are increasingly attracting guests from the Czech Republic, Slovakia, Poland.

A simple rule for getting started: Add the languages of neighbouring countries and the nations that most often travel to your region — plus English as a universal option. For most European destinations, that's 3–5 languages.

An important nuance: English is the common default language, but not a substitute for the native language. CSA Research notes that even among speakers of languages with a high level of English proficiency, most prefer to read commercial information in their native language. An English version of the menu is needed as a universal option. German, Polish, French — as respect for the specific guest.

Digital vs print: what the data says

The conversation about digital menus often comes down to technology hype. But behind it lies quite prosaic economics.

Printing costs: numbers that are inconvenient to ignore

According to industry analysis by Evergreen HQ (aggregating operator data and WAND Corporation), a restaurant updating its menu four times a year at a print run cost of $500–2,000 spends $2,000–8,000 annually on printing — and that's just one language. Five languages multiply this figure proportionally.

A digital menu eliminates these costs entirely. A QR platform subscription costs tens of times less. Updates — instant, free of charge.

Revenue: what McDonald's says

McDonald's — the world's largest restaurant format operator — publicly reported that the transition to digital menus and multilingual kiosks yielded an increase in average spend of 3–5% at American locations. For a single restaurant with a turnover of €500,000 per year, that's €15,000–25,000 in additional revenue from just one change in how the menu is presented.

The mechanism is clear: when a guest understands what's on offer, they order more fully. No anxiety from unfamiliar names — no defensive minimalism in ordering.

Three typical scenarios: how it works in practice

Below are illustrative scenarios based on typical European restaurant profiles. These are not real cases with establishment names, but generalised pictures of what happens when switching to a multilingual menu in different contexts.

Language switching UX: the details that decide everything

You can make the right translation and the wrong interface — and lose half the effect. The UX of a language switcher is a detail that often goes unthought-through.

Flags vs text labels

Using flags to select a language is intuitively understandable but a poor solution. The Swiss flag doesn't mean 'German' — Switzerland has four official languages. The UK flag is not 'English' for an Australian, an Irishman, or an American. The Belgian flag won't tell you whether it's Dutch or French.

The right solution is text labels in the language itself: Deutsch, Français, Polski, Čeština. The guest sees their word and clicks without hesitation. This is universal and requires no knowledge of European flag geography.

Auto-detection vs manual selection

Auto-detecting the language by device locale is a good starting point. The browser passes device language information, and the platform automatically shows the right version. The guest doesn't think about switching at all.

But auto-detection is a hypothesis, not a fact. A German phone may belong to a Dutch person. Therefore, a manual switcher must always be accessible and visible — not hidden at the bottom of the page.

Sticky preference: a menu that remembers the guest

If a guest has manually selected a language, a good solution saves that choice on the device. When re-scanning the QR code — even after a few days — the menu opens immediately in the right language. This is a small detail, but it signals: 'we remember you'.

EU Regulation 1169/2011

What to do: from zero to a working menu

Everything discussed above has stopped being complex or expensive. The infrastructure that previously required an agency, a budget and weeks of development is now set up in a matter of minutes. The algorithm is simple:

After two to three weeks, scan analytics will show the real language profile of your guests — and you'll be able to add or remove languages based on fact, not assumptions.

Menu updates — seasonal items, price changes, temporary offers — are reflected instantly, without reprinting, without delay, in all languages simultaneously. The QR code remains the same: the guest doesn't need to relearn anything.

FAQ