---
title: "Restaurant chatbot on a table QR code menu: suggest dishes, grow guest loyalty, get feedback"
description: "A QR code on the table can open a restaurant chatbot that knows your menu: it asks guests their taste, diet and allergies, suggests dishes you serve, then grows your email list."
canonical: "https://linkbreakers.com/help/article/how-restaurants-use-qr-codes-for-table-ordering-guest-feedback-and-loyalty"
last-updated: 2026-11-15
---

# Restaurant chatbot on a table QR code menu: suggest dishes, grow guest loyalty, get feedback

> A QR code on the table can open a restaurant chatbot that knows your menu: it asks guests their taste, diet and allergies, suggests dishes you serve, then grows your email list.

Video: [How to add an AI chatbot to your restaurant's QR code menu](https://www.youtube.com/watch?v=2jVqHf13JhM)

*The film: setting up the menu assistant and a real chat after a guest scans the table's QR code, recorded in our demo workspace on 15 November 2026. The restaurant, Fernhill Kitchen, its agent Sage and its menu are made up.*

## Short answer

A restaurant chatbot on a QR code menu is a chat that opens when a guest scans the code on their table. An AI agent that knows your menu asks what they feel like, their diet, their allergies and how spicy they like their food, one question at a time. It suggests dishes from your menu, and says so when a dish is not on it: asked for a veggie pizza, ours answered "We do not have pizza on our menu". Each answer is saved on the guest's visitor profile as it comes.

Then the visit goes on. In our demo, a short form invites the guest to join the restaurant's club, with an email field and a marketing consent box, and then the full menu opens. A guest who only wants the menu taps "Just show me the menu" and gets it at once.

In Linkbreakers, the chat is a **[Chat bot](/workflow-steps/chat-bot)** step run by one of your **agents**, on every plan, including Free. Each message the agent answers uses platform credits, based on what the AI model used for it, never more than 100 for one message (what our demo spent is under the limits below). What you get: a guest who cannot choose can find a dish that fits them in a few messages, you learn what your guests like and avoid, and those who tick the box are on your email list, with their consent recorded, before they leave the table.

:::try-ai-prompt {"prompt":"Create a friendly Product expert agent called Sage for our restaurant, with Only answer from knowledge on, that suggests one or two dishes from our menu once it knows the guest's mood, diet and allergies, and reminds guests with an allergy to tell their server. Then add a Chat bot step right after the scan on our table QR code link: Sage learns mood, diet (None, Vegetarian, Vegan, Gluten-free), allergies, spice level (Mild, Medium, Hot) and party size, all required, with a skip button labelled Just show me the menu. After Completed, add a form called Join our club with an email field and a marketing consent box, then open our menu from Media. Send Not completed straight to the menu.","description":"Ask the Assistant to build the agent and the table chat, then paste your menu into the agent's knowledge.","target":"assistant"}:::

## How a restaurant chatbot on a QR code menu works

A QR code is a link printed as a pattern of squares. When a guest scans the code on a table tent, their phone opens a web page. Here, that page is a chat. There is nothing to install.

The chat is run by an AI agent: an assistant you brief once, with its job, its tone and what it knows. Here, it knows your menu, with each dish's diet, spice level and allergens. It asks the guest what you want to know, answers their questions from the menu, and suggests the dishes that match their answers. Each answer is saved. When the chat ends, the next step of the visit opens: a form, a page or the menu itself.

A printed menu shows every guest the same list. The chat starts from what this guest likes and avoids, and points to the dishes on your list that fit.

## Set up a restaurant QR code chatbot, step by step

This is the setup we built and tested in our demo workspace on 15 November 2026, for a made-up restaurant, Fernhill Kitchen.

### 1. Create the agent and paste your menu

Open **Agents** in the dashboard menu and click **New agent**. Ours is called Sage, with these settings:

- **Role**: Product expert, which "presents the product and answers questions about it". Here, the product is your menu.
- **Tone**: Friendly, with **Reply length** set to Short, the guest's own language and emoji off.
- **Visitor questions**: Answer, then steer back. A guest who asks something gets an answer, and the agent goes back to what it needs to learn.
- **Instructions**: who the restaurant is, that guests scan the code on their table and order with their server, and when to suggest dishes. Ours: suggest one or two dishes by their name on the menu once you know the guest's mood, diet and allergies; only suggest dishes whose listed allergens do not include the guest's allergy, and remind them to tell their server about it; never say the kitchen can change or leave out an ingredient.
- **Guardrails**: Stay on topic, **Only answer from knowledge** (it says it does not know rather than inventing a dish), No promises, No professional advice and Decline offensive requests.
- **Knowledge**: one Markdown document, "Dinner menu", with an entry per dish and the kitchen's note on allergies. An agent can hold up to 20 documents and 50,000 characters; our 14 dishes take about 2,550.

One dish from our menu document:

```markdown
### Green curry with tofu, $18
- What it is: Thai green curry, aubergine, green beans, jasmine rice
- Diet: Vegan, Vegetarian, Gluten-free
- Spice level: Hot
- Allergens: soy
```

Under its own heading, the document also says what the kitchen handles and that guests with an allergy should tell their server before ordering. Every setting is described in [how to create an agent](https://linkbreakers.com/help/article/how-to-create-an-agent).

### 2. Add a Chat bot step after the scan

Open the table code's link (its workflow opens), click **+** on the connector right after the scan, and choose **Chat Bot**. Pick the agent and write an opening message, for example "Hi! I'm Sage, Fernhill Kitchen's menu assistant. What are you in the mood for tonight?". Under **What the agent should learn**, list each item with a label, the key it is saved as, a type and a short explanation the agent reads:

| Label | Saved as | Type | Required |
|---|---|---|---|
| Mood | `mood` | Text | Yes |
| Diet | `diet` | One of a few choices: None, Vegetarian, Vegan, Gluten-free | Yes |
| Allergies | `allergies` | Text | Yes |
| Spice level | `spice_level` | One of a few choices: Mild, Medium, Hot | Yes |
| Party size | `party_size` | Number | Yes |

A step holds up to 10 items, and the chat counts as **Completed** only when every required item is known. Turn on **Show a skip button** and label it "Just show me the menu", so a guest who already knows what they want is never stuck in the chat. Guests can send 15 messages by default, and you can set anything from 3 to 30. The other settings are explained in [how to add an AI chat bot to a QR code](https://linkbreakers.com/help/article/how-to-add-an-ai-chat-bot-to-a-qr-code).

### 3. After the chat: the club form, then the full menu

On the chat's **Completed** branch, add a **[Form](/workflow-steps/form)**. Ours is titled "Join the Fernhill club", with one Email field and a marketing consent box in our own words: "Yes, send me news and offers by email." The box records each guest's yes or no with proof, and connected email tools subscribe only those who said yes. A "No thanks" skip button keeps joining a choice.

After the form, add an **[Open media](/workflow-steps/open-media)** step that hands out your full menu from **Media**, as an image or a PDF. Then connect the chat's **Not completed** branch to the same Open media step, so a guest who taps the skip button, or whose chat stops early, still gets the menu.

### 4. Print the code and test it from a phone

Download the QR code as SVG or high-resolution PNG for your table tents. Before printing, scan it with your phone and go down each path once: a full chat, the club form, and the skip button.

## A restaurant chatbot example: what the guest sees

In our test from an emulated iPhone, the chat opened in the restaurant's page theme. The guest answered "Something spicy.", "Vegetarian." and "I'm allergic to peanuts.". The agent asked how spicy they like their food and reminded them to tell their server about the allergy.

The guest then asked "Do you have a veggie pizza?". The menu has no pizza, and the agent said so: "We do not have pizza on our menu, but how spicy do you like your food?". After "Hot, please.", it suggested the crispy cauliflower bites or the green curry with tofu. On the menu, both are vegetarian and hot, and neither lists peanuts. It reminded the guest again to tell their server, asked how many people were eating, and closed with "Thank you so much." after "Two of us.": six messages from the guest in all.

The club form opened. The guest typed an email, ticked the box and tapped **Join**, and the full menu opened. In a second test, a tap on "Just show me the menu" went straight to the menu.

Where you find it afterwards:

- **The visitor profile**, in **Visitors**: the email and the guest's marketing consent, and each chat answer under **Custom Attributes**.
- **The visit's activity timeline**: the chat with each answer and the full conversation, and the menu being opened.
- **Team notifications and webhooks**, which include the chat's outcome and answers, so they can reach your team or your own tools.

After service, filter **Visitors** by an attribute, for example `diet` equals `vegetarian`, to see how many of the guests who chatted are vegetarian. The conversations show what guests asked for that you do not serve.

## More ways restaurants use a table QR code

### QR code menu for restaurants: change it without reprinting

A QR code at the table can also open the menu directly, with no chat. An Open media step hands out your menu as a PDF or an image, and **Replace file** in Media puts next week's menu in its place while the printed code stays the same ([how to create a QR code for a PDF](https://linkbreakers.com/help/article/how-to-create-a-qr-code-for-a-pdf)). A [time of day condition](https://linkbreakers.com/help/article/how-to-use-scheduler-conditions) can open the lunch menu at noon and the dinner menu in the evening.

To know which table a scan came from, you do not need one link per table. On a custom domain, one [dynamic short link](/help/article/dynamic-slugs-capture-data-from-the-url-path) such as `menu/{table}` prints as one QR code per table, and each scan saves the table number on the visitor, next to their chat answers. If you run several restaurants, [tags](https://linkbreakers.com/help/article/how-to-use-tags-to-segment-and-analyze-qr-campaigns) group each venue's codes so you can compare them.

### Guest feedback, and reviews from every guest

A code on the bill or on a table tent at the end of the meal can open a short [feedback form](https://linkbreakers.com/help/article/how-to-create-nps-and-product-feedback-forms) with a 0 to 10 rating and a comment box. A [Visitor data condition](/workflow-steps/visitor-data-condition) reads the score, so a low score can lead to a private "tell us what went wrong" form, and your team can hear about it while the guest is still in the room. If you prefer a conversation, an agent with the Feedback collector role can ask the same questions in a chat.

Show your public review link to every guest. Asking only happy guests for a review is called review gating, and the largest review platforms forbid it. The score can change what comes first, not who may leave a review: [how to route visitors based on NPS scores](https://linkbreakers.com/help/article/how-to-route-visitors-based-on-nps-scores) shows the build that follows those rules.

### Loyalty club and email list

The club form after the chat is a loyalty sign-up: an email and a consent box. You can also put the same form behind a code on the bill, the receipt or the host stand. Give each placement its own link, and you see which one brings the most scans and sign-ups. The guests who said yes are the ones you can email about a new menu or a quiet Tuesday; [how coffee shops use QR codes for loyalty programs](https://linkbreakers.com/help/article/how-coffee-shops-and-cafes-use-qr-codes-for-loyalty-programs) has more on rewards and repeat visits.

## Limits and caveats

- **The agent only knows what you paste in.** It has no link to your kitchen, your stock or your till. A dish that sold out tonight, or a recipe that changed, stays on its menu until you update the knowledge document.
- **Allergies: the agent repeats your menu, it does not check the food.** It can say which dishes list peanuts, because you wrote it. It cannot know about cross-contact in the kitchen or a supplier's change. Put your kitchen's allergy note in the knowledge, tell the agent to remind guests with an allergy to tell their server, as ours did, and never present the chat as allergy safety advice. Your staff stay the ones who answer allergy questions.
- **A guardrail is an instruction to the model, not a guarantee.** Only answer from knowledge, No promises and the others are rules the AI model is told to follow. They steer it; they cannot make a mistake impossible. Test the chat with the questions your guests ask, and read conversations in the activity timeline.
- **Suggestions are said in the chat, not saved as data.** The profile keeps the guest's answers, not the dishes the agent suggested. To see the suggestions, open the conversation.
- **A choice answer is one value.** A guest who is vegetarian and gluten-free is saved with one diet. Use a Text item when one answer may hold several.
- **Make required what you need asked.** The chat completes as soon as every required item is known. In our first tests, with spice level and party size optional, the agent finished without asking them.
- **Messages use platform credits.** Based on what the AI model used, never more than 100 a message; a longer menu in the knowledge costs more per message (ours, 2,550 characters: 110 credits for three messages on 15 November 2026) ([what the chat costs](https://linkbreakers.com/help/article/how-to-add-an-ai-chat-bot-to-a-qr-code#what-it-costs)). When the workspace runs out, the chat is skipped and guests go on along **Not completed**, straight to your menu, so the code keeps working. The same branch catches a chat that stops because the model failed.
- **The chat does not take orders.** Guests order with their server, or on your own ordering page, which a Redirect step can open after the chat.
- **Consent is recorded for email only.** The box is about your emails, nothing else.
- **The chat and the menu need a connection.** Where mobile signal is weak, offer guest WiFi or keep a few printed menus.

## Frequently asked questions

### Can a QR code menu have an AI chatbot?

Yes. The QR code opens a short link, and the link can show a chat before the menu. In Linkbreakers, add a **Chat bot** step right after the scan, with an agent that has your menu as its knowledge. The guest chats in your page theme, with no app to install, and the menu opens when the chat ends.

### What happens when a guest asks for a dish you do not serve?

With **Only answer from knowledge** on, the agent says it is not on the menu and goes back to what it needs to learn. In our test, asked for a veggie pizza, it answered "We do not have pizza on our menu" and later suggested two vegetarian dishes from the menu.

### Can the chatbot answer allergy questions?

Only by repeating what your menu says: the allergens you listed for each dish. It cannot check the food or the kitchen. Tell it to remind guests with an allergy to tell their server, and never present the chat as allergy safety advice.

### Can guests skip the chat and just see the menu?

Yes. With **Show a skip button** on, a guest taps it, for example "Just show me the menu", and goes on along **Not completed**. Connect that branch to the menu, and nobody has to chat to see it.

### How much does a restaurant chatbot on a QR code cost?

The Chat bot step is on every Linkbreakers plan, including Free. Each message the agent answers costs platform credits (never more than 100; about 37 a message in our demo), and the Free plan includes 5,000 credits a month ([pricing](https://linkbreakers.com/pricing)). Messages are charged to your workspace, never to the guest.

## Next steps

1. [Add a Chat bot step to your table's QR code](https://linkbreakers.com/help/article/how-to-add-an-ai-chat-bot-to-a-qr-code), with the questions your servers ask most.
2. [Create the agent](https://linkbreakers.com/help/article/how-to-create-an-agent) and paste your menu, with each dish's diet and allergens, into its knowledge.
3. [Put your full menu behind the code](https://linkbreakers.com/help/article/how-to-share-media-from-a-qr-code) with an Open media step, and replace it when the menu changes.

## Sources

- Our own test in the Linkbreakers demo workspace, 15 November 2026: the agent, its menu knowledge, the Chat bot step, the club form and the Open media step described above, and chats recorded from emulated iPhones. The screenshot at the top is from that test.
- Google, Maps user generated content policy, "Prohibited & restricted content" (selectively soliciting positive reviews): https://support.google.com/contributionpolicy/answer/7400114
- Linkbreakers help: [how to add an AI chat bot to a QR code](https://linkbreakers.com/help/article/how-to-add-an-ai-chat-bot-to-a-qr-code) (branches, credits, where the answers go) and [how to create an agent](https://linkbreakers.com/help/article/how-to-create-an-agent) (roles, guardrails, knowledge limits).

## Last reviewed

15 November 2026, by Laurent Schaffner, against the live product.
