Key takeaways
- Website chat is good at hours, directions, parking, policies and pointing to booking and ordering links.
- It should hand off bookings to your reservation system and allergy questions to staff, never confirm or guess.
- Many guests are wary of AI[1]; say clearly that it is automated and offer a person.
- Review transcripts weekly, fix the source information and track wrong answers and handoffs.
What a website chat is for
A chat window on a restaurant website answers questions guests would otherwise call about or give up on: whether you are open tonight, where to park, whether the patio is heated, whether you take reservations for ten. When it works, it saves guests a phone call during your busiest hours and catches visitors who would not call at all. When it fails, it gives wrong hours with total confidence, covers your menu with a pop-up and annoys guests who just wanted the booking button.
Phones still dominate for many guests, and those who prefer a person are not a small group: in a Cornell study, 65 percent of diners who had not made an online reservation said they wanted to talk to someone to book[2]. Chat is an extra door, not a replacement for the phone or the booking page.
What chat can and cannot answer
| Question type | Chat should | Source of the answer |
|---|---|---|
| Hours, holiday closures, kitchen close | Answer directly | Dated house information document |
| Location, parking, transit, entrances | Answer directly, with a map link | House information document |
| Reservations | Link to the booking page, or take a request for staff to confirm | Reservation system or request form |
| Large parties and private events | Collect date, party size and contact details for the events manager | Event request fields |
| Menu questions | Describe dishes from the current menu and link to it | Current menu text |
| Allergies and dietary safety | Never answer safety questions; pass to staff with a callback | Handoff rule |
| Complaints | Apologize briefly and pass to a manager with contact details | Handoff rule |
| Prices and offers | Quote only what is on the current menu or offer page | Menu and offer pages |
| Order status | Point to the order confirmation or the ordering platform's help | Ordering platform |
The rule underneath the table is simple: chat answers facts it has been given and routes everything else. It should be easy for a guest to see when they have moved from an answer to a request.
Handing off bookings properly
Booking is where chat most often goes wrong. A chat that says "you're booked for 7:30" when it cannot see the reservation book creates a no-table moment at the host stand. Pick one of these patterns and stick to it.
Link to the booking page
For standard reservations, the chat offers the booking link with the date and party size already filled in, if your reservation system supports it.
Take a request
If the booking page cannot handle the request, such as a large party or a special occasion, chat collects name, phone, email, date, time, party size and notes, and says clearly: "This is a request. Our team will confirm by phone or email."
Say when to expect an answer
Give a real time frame: "We reply to requests by noon the next day."
Send it where someone will see it
Requests go to a shared inbox or the reservation system's request queue, not to an owner's personal email.
Confirm in writing
Staff confirm the booking through the reservation system so the guest gets the normal confirmation.
Guests are flexible about channels when they want a table: in the Cornell research, 58 percent of diners said they would still try to book through another channel if a restaurant was not on their preferred reservation site, and only 18 percent said they would not book[3]. A chat that offers a clear next step catches the first group.
Source: Cornell University Center for Hospitality Research (Kimes and Kies), 2012[3]
Show the data
| Item | Value |
|---|---|
| Would still book another way | 58% |
| Would not book | 18% |
Disclosure, trust and privacy
Many guests are cautious about AI. Fifty percent of US adults say they are more concerned than excited about its increased use in daily life[1], about six in ten would like more control over how AI is used in their own lives[4], and 76 percent say it is extremely or very important to be able to tell whether text was made by AI or by people[5].
- Label the chat as an automated assistant in the opening line, not in small print.
- Offer a person: the phone number, a callback request or the hours when staff answer messages.
- Collect only what you need for the request: name, contact details, date, party size and notes.
- Never ask for card numbers in chat. Deposits go through your reservation or payment system.
- State in your privacy notice that chat conversations are stored and what they are used for.
If the chat cannot answer, it should say so plainly and offer the next step. "I'm not sure, here is how to reach the team" is better than a confident guess.
Lines to write before launch
The chat's own wording sets expectations. Write these lines yourself rather than accepting the vendor's defaults, and keep them in the house information document.
- Opening: "Hi, I'm the automated assistant for [restaurant]. I can help with hours, directions, menus and booking. For anything else, I'll pass you to the team."
- Booking: "You can book here with your date and party size filled in. For 9 or more, I can send a request to our events manager."
- Request confirmation: "Thanks. This is a request, not a booking yet. The team will reply by phone or email by noon tomorrow."
- Allergy: "For allergies and dietary safety, please speak with our staff directly. Can I take your name and number so a manager can call you?"
- Complaint: "I'm sorry to hear that. I'll pass this to the manager. What is the best number or email to reach you?"
- Unknown: "I don't have that information. You can call us at [number] during opening hours, or leave your details and we'll get back to you."
Out-of-hours questions and event leads
A chat earns much of its value when nobody is there to answer. Late-night visitors planning tomorrow's dinner, office managers organizing a team lunch and families planning a celebration often visit the website after hours. Make sure out-of-hours requests are clearly time-stamped, reach a shared inbox that is checked first thing, and get a reply the same morning. Event and large-party requests are often worth more than a whole night's walk-ins, so route them to the person who quotes events, not to a general inbox.
Design it so it does not get in the way
- Load the chat after the page and menu, so it does not slow the first screen on phones.
- Show a small button, not a window that opens over the menu or the booking button.
- Do not auto-open with a greeting on every page; guests looking for the menu will close it and may leave.
- Keep the booking, ordering and call buttons visible on the page whatever the chat does.
- Test on a mid-range phone over cellular data. Sixteen percent of US adults are smartphone-only internet users[6], and a heavy chat widget hurts them most.
Check accuracy every week
Chat accuracy decays as menus, hours and policies change. Build a short review into the manager's week.
- Read a sample of the week's conversations, including every one that ended in a handoff or a complaint.
- Mark each wrong or weak answer and find its source: missing information, old information or a misunderstood question.
- Fix the source information first, then the chat settings.
- Rerun a fixed list of test questions after any change: holiday hours, large-party booking, dogs on the patio, kitchen close, allergy question, complaint.
- Track four numbers monthly: conversations, booking or event requests captured, handoffs to staff, and wrong answers found.
Wrong answers should trend toward zero. If they do not, the information behind the chat is not being maintained, and the chat is doing more harm than good.
When chat is worth it, and when it is not
- Worth it: busy restaurants with frequent repeat questions, many large-party or event requests, or a lot of out-of-hours website traffic.
- Worth it: restaurants whose phones ring out at peak and who already keep their house information current.
- Not yet: restaurants without a reliable, dated source for hours and policies.
- Not yet: very small restaurants where a clear website answers almost everything.
Example: a fictional 110-seat restaurant with a private dining room adds website chat. In the first month it answers mostly parking and hours questions and captures 18 event requests, which the events manager calls back the next morning. The weekly review finds it describing the patio as open in November. The manager adds dated seasonal notes to the house information document, and the Example wrong-answer count falls to zero the following month.
Frequently asked questions
Can a website chat take restaurant reservations?
It should link to your booking page or take a request for staff to confirm. It should never confirm a table it cannot see in the reservation system.
Should restaurant chat answer allergy questions?
No. It should pass allergy and dietary-safety questions to staff with the guest's contact details.
Do guests like AI chat on restaurant websites?
Many are cautious[1], so label it as automated, keep answers accurate and offer a person.
How do I know if my website chat is accurate?
Read a sample of conversations weekly, rerun a fixed list of test questions after changes, and track wrong answers and handoffs.
References
Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) ↑
Cornell University Center for Hospitality Research (Kimes and Kies) (2012). The Role of Multi-Restaurant Reservation Sites in Restaurant Distribution Management (Cornell Hospitality Report, 2012) (opens in a new tab) ↑
Cornell University Center for Hospitality Research (Kimes and Kies) (2012). The Role of Multi-Restaurant Reservation Sites in Restaurant Distribution Management (Cornell Hospitality Report, 2012) (opens in a new tab) ↑
Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) ↑
Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) ↑
Pew Research Center (2026). Americans’ Use of Mobile Devices in 2026 (Fact Sheet) (opens in a new tab) ↑