The table you lost last night was booked by a website that couldn't talk back.

An AI receptionist for a restaurant answers the questions diners ask before they commit — are you open tonight, table for four, do you do gluten-free, is there parking — the moment they ask, so a hungry diner doesn't hit a dead form and book the place down the road. Plenty of that still happens by phone. Maya covers the other front door: the website, by voice or text, 24/7, in the diner's own language, reading your own menu and pages so the answers are accurate.

Last updated 2026-07-16

Where the phone stops

Restaurants do live on the phone — but the phone only rings during a shift, and only for people who call. The diner deciding at 11pm, the tourist who doesn't want to phone in a language they're unsure of, the couple checking whether you do gluten-free before they leave the house — they're on the website, and the website answers none of them. That's the front door nobody's staffing, and it's where covers quietly leak.

Maya covers that door. Once an assistant can greet the diner in their own language, answer the menu and availability questions from your real content, capture the booking, and hand it to you framed and ready — "receptionist" is the small version of what it does. It's the layer that turns your website's after-hours and tourist traffic into covered tables, without adding a host.

What Maya does for a restaurant

It answers the shift-eating questions the moment they're asked — at any hour — hours, availability, the menu, dietary and allergen options, location, parking — answered by voice or text from your own menu and pages, at 11pm while your kitchen is slammed and nobody's near the phone. The diner who'd have given up gets a real answer and books.

It greets each diner in their own language — the tourist moment — a tourist arrives on your English site with their phone set to French, and Maya greets and answers in French, no selector clicked. In a tourist area that one moment — met in their own language — is the difference between a booking and a bounce.

It follows the diner's language even mid-conversation — a diner can start in English and switch to German mid-question — Maya follows without missing a beat and without losing what they asked, and switches to a language version your site already has. The foreign table that couldn't navigate your site still books.

It captures the booking framed and ready — who, how many, when, plus "one guest is coeliac, asked about the tasting menu" — captured in the conversation and handed to you ready to confirm, or booked directly when your calendar is connected. (Booking plan and up; even the entry plan captures the intent so nothing is lost.)

It asks happy diners for a review, within the rules — after a good interaction it can invite a review, gated so it only asks genuinely satisfied guests — the reputation flywheel that fills tables, without a host having to chase it.

It tells you what diners kept asking for that isn't clear on your site — "nine diners this week asked whether you take large groups" — your own traffic surfaces the missing info or the service worth adding.

What happens when Maya doesn't know the answer?

This is the question every owner asks about AI — especially where allergens are involved — and it's where Maya is careful. Maya checks its answers against your approved menu and site content before it responds. When it can't verify something — whether a dish is truly gluten-free, whether you can seat a group of twelve — it says so instead of guessing, offers the diner the right next step (often "let me pass this to the team"), and flags it for you.

What that means for you: Maya never guesses an allergen answer or invents availability to satisfy a confident question — on a dietary question, a made-up answer is a genuine safety and trust problem, so it defers to your team instead. Honest limit: it answers from what your menu and pages actually say; anything not there, or anything that needs a human judgement call, it hands off rather than inventing.