Lulu Chat Policy

You’re talking to an AI

Lulu is an artificial intelligence: a large language model, not a human travel agent, and not a Chinerary staff member reading along. No one at Chinerary sees your conversation in real time, and replies are generated automatically, not written by a person.

What happens to your messages

When you send Lulu a message, it’s sent (along with relevant context from your own itinerary and budget, so she can give a useful answer instead of a generic one) to OpenRouter, a third-party AI routing service based in the United States, and from there to the underlying AI model that actually generates the reply. That’s the only place your message goes.

Where that AI model is actually hosted

Which model is configured for Lulu can change over time, and has included both US- and Chinese-developed models. Whichever one is active, OpenRouter can serve it through more than one underlying infrastructure provider, in more than one country; we don’t control that choice directly, only which model is selected. Every request also carries one technical restriction regardless of which model or provider is used: never use a provider that would store the message or train on it.

We don’t restrict which country actually processes the message beyond that: a message you send Lulu can genuinely be processed outside the EU, including in the US or China, depending on the model in use at the time. Because of that, the first time you use Lulu’s chat, we ask you to explicitly confirm you understand and accept this before your first message is sent; that consent screen is the actual safeguard here, not a technical block on any particular country. If you’d rather not have a message processed outside the EU at all, the reliable thing to do is not use the chat (or not send that particular message). Declining doesn’t affect anything else in the app.

By default, we don’t keep a conversation history for you: the chat lives in your browser for that session and is gone when you refresh the page or start a new chat, and there’s no way to reopen it.

What we do log, for every message: each request to the AI is written to a technical log that only our administrators can view. We use it to find out why a reply failed, and what makes a request slow or expensive. It has two parts:

• Usage metadata: which model was used, token counts, cost, how long it took, whether it succeeded, and the error if it didn’t. Kept for as long as your account exists.
• The full request and response: your message and the earlier messages in that chat, the itinerary and budget context sent with it, the results of any lookups Lulu made (weather, rail connections, and so on), and Lulu’s reply. Kept for 30 days, then automatically deleted.

Both are deleted along with your account. This log is separate from conversation history below: it isn’t something you can browse, and turning history off doesn’t turn it off.

If you turn on “Save Lulu conversations” (a toggle in Settings, off by default), that changes: your messages and Lulu’s replies are saved to your account so you can revisit or continue a past conversation from the conversation history page, and they stay saved until you delete them yourself, either one conversation at a time or all at once. Turning the setting back off stops any new messages from being saved; it doesn’t retroactively delete what was already saved while it was on.

Separately from that setting: if you use the Report button under one of Lulu’s replies, we save that specific reply and a little surrounding conversation so we can see what went wrong and improve how Lulu answers, regardless of whether conversation history is turned on. This only happens when you press that button yourself, never automatically.

Don’t share sensitive information

Please don’t paste passport numbers, financial account details, health information, or other sensitive personal data into the chat. Lulu doesn’t need it to help you plan a trip, and every message still passes through a third-party AI provider to generate a reply regardless of your history setting; the same caution you’d apply with any AI chat tool applies here.

Lulu can be wrong

Like any AI, Lulu can state something incorrectly and sound just as confident doing it. Treat her answers as a starting point (especially anything involving visas, customs rules, current prices, or train schedules) and verify what actually matters against an official or current source before relying on it.

What Lulu can see and do

To give relevant answers, Lulu’s replies are grounded in your actual trip: your route, trip dates, and (if you’ve filled it in) your travel preferences from the onboarding survey. On a Premium account, she can also call a small set of tools (for example, looking up the weather forecast for a city on your route) which send that city’s name to our weather provider, not any personal data about you.

More detail

This page covers Lulu specifically. For how we handle data more broadly, including your rights to access, export, or delete your account, see the full Privacy Policy.