AI notice

Version 1.1.2Effective 6 October 2026Last updated: 14 October 2026

The concierge and the trip planner run on language models operated by third parties. Here is what gets sent, how we ask your permission before sending it, who keeps it, and what you must not use it for.

1. Who processes your messages

The concierge — the conversation — runs on a model from Google (Gemini).
The trip planner — the one that builds the itinerary — runs on a model from Google (Gemini).

Both operate from the United States, so using these features involves an international transfer of data.

2. What is sent, and what is not

What is sent is your message, a description of the destination, and the rows from our own catalogue the model looked at to answer you (a beach’s details, a restaurant’s opening hours).

Your trip is sent too. If you have a saved trip, the concierge can look it up to answer you, and the planner uses it to build or adjust your itinerary: the dates, how many of you are travelling and the children’s ages, your budget and interests, your flights, your accommodation’s name and dates (to the planner, its address as well), how you get around, your trip notes, your itinerary and, if you wrote them, your dietary, accessibility or special-occasion notes. Your accommodation’s confirmation number and your notes about it are not sent.

Your identity is not sent. The text we prepare for the model contains no name, email address, username or account identifier. The provider receives a conversation; it does not receive whose it is.

That stops being true if you type personal data into the message or into your trip notes. Everything you type travels verbatim. Do not put anything into the concierge that you would not want a third party to process: card numbers, passwords, health details, identity documents, or other people’s personal data.

3. Your permission, before anything is sent

Before your first concierge message, the first itinerary the planner builds or the first AI change to a day goes to Google, we show you a notice that says what is sent — what you type and your trip details — and to whom, and asks your permission. Nothing is sent unless you tap “Continue”. If you tap “Not now”, we send nothing, we do not record that answer, and we ask again the next time you try to use the AI; the rest of the app works the same.

If you give permission, we record it on our servers, linked to your account: which version of this notice it was, when, in which language and from which platform. Your device also keeps a copy so that we do not ask on every use. Our servers refuse concierge messages, planner itineraries and AI changes to a day from an account without that record, before anything is sent to Google. If this notice changes materially, we ask you again.

This permission covers what you send as a traveller. The writing assistant for businesses — “Write with AI”, in the business dashboard — does not ask for it, because it sends nothing from a traveller: it sends Google the business listing’s details, its current description and, for a promotion, what the owner wrote about it, and only when the owner chooses to use it and taps “Create draft” or “Another draft”. The translator for businesses — the “Translate” button, in the same dashboard — does not ask for it either: it sends Google the texts the owner chooses to translate (their description, menu, an event, an offer, their tagline or their confirmation message) and the business’s name, and only when the owner taps “Translate”; the translation comes back as a draft, and nothing is published until the owner reviews it and taps “Save”. The same goes for the assistant our team uses to draft listings.

The AI works only with an account. On an accommodation’s guest tablet it is neither asked for nor used: there we ask you to open it on your phone.

You can withdraw the permission whenever you like, by asking through Your data and your rights; from then on we do not send your data to the AI until you give it again. There is no button to withdraw it inside the app today.

4. Who keeps the conversations

Google keeps a copy of the conversation on its servers. The API we use stores each exchange on the provider’s side so that a follow-up turn does not have to resend the whole history. According to Google’s documentation for the paid service we use, those stored conversations are kept for 55 days; separately, Google logs prompts and replies for a limited time solely to detect and prevent abuse of its usage policy, and does not use them to improve its products. In other words: as well as our copy, there is a copy at Google.

We keep a copy too — your messages and the model’s, linked to your account — so the conversation is still there when you come back.

Our copy has no automatic deletion deadline today. It is a known gap and we say so rather than leave it implied. You can remove it by deleting your account or by asking for that data to be cancelled.

Deleting your account does not delete Google’s copy at that moment. It removes ours straight away; Google’s goes when its period runs out. If it matters to you that nothing is left at the provider even for that long, the realistic answer is not to use the concierge or the planner for anything you would not want stored beyond our reach.

5. It can be wrong

A language model produces the most likely text, not the truest. It can invent an opening time, confuse two beaches, quote a price that has changed, or state something false with complete confidence. When it answers from our catalogue it also inherits the catalogue’s mistakes.

Always confirm anything that matters — an opening time, a price, whether a road is passable, whether a tour is running today — directly with whoever provides it.

6. What it is NOT a substitute for

The concierge is a travel assistant. It is never a substitute for:

emergency services — in Mexico, 911;
a doctor, for diagnosis or for treatment;
a lawyer, an accountant or any professional advice;
an authority — immigration, customs, police, civil protection — on procedures, requirements or permits;
the operator of a tour, transport service or hotel on their own terms, prices and availability;
official weather warnings from the Servicio Meteorológico Nacional and Protección Civil;
the maritime authority and the beach flags on sea state, currents and whether it is safe to go in.

In an emergency, put the app down and call 911.

7. It decides nothing about you

We do not use these models to make automated decisions that produce legal effects for you or that significantly affect you: they do not assess your creditworthiness, your behaviour or your access to the service. They suggest places and assemble itineraries.

8. Limits on use

Concierge use is capped per account per day, because every reply costs money. When you reach the cap we tell you, and the rest of the app carries on working normally.

You may not use the concierge to generate unlawful content, to extract our catalogue in bulk, or to automate requests. See the Terms of Service.