How to share trip details with AI without oversharing

Planning a transfer usually needs an arrival time and airport, not a passport scan. Share the minimum information needed for the task and keep credentials and unnecessary personal details out of reusable prompts.

In this guide: Use only the trip information needed for the planning task.

1. Identify the planning fields

For a schedule check, dates, times, locations and relevant booking conditions may be enough. Use traveller labels where names are unnecessary. A room-layout question does not need someone’s home address.

2. Remove unrelated sensitive details

Before pasting a confirmation, review it for payment information, access links, personal identifiers and details about other travellers. Use the provider’s current privacy controls and policies; do not assume every account has identical settings.

3. Review connected actions separately

Reading a supplied summary and changing a saved trip are different tasks. Name the correct trip, preview the exact change and approve only what you intend. Never include sign-in codes or tokens in a public example.

Worked example: what this looks like

Fictional scenario—not a real customer story or executed AI result.

Simulated example: a traveller wants to check whether dinner fits after arrival. They provide ‘Traveller A, arrive [airport] 4 pm, hotel [area], dinner 7 pm’ rather than uploading an unredacted travel document. The assistant can analyse the timing and ask for transfer assumptions without seeing unrelated identity information.

Give the assistant a test it can actually fail

Define what a useful answer must preserve before requesting it. Examples include the same traveller count, a fixed arrival date, no car hire, a stated budget scope or an unchanged confirmed booking. Then read the answer against those requirements instead of judging its confidence.

When something fails, describe the failure precisely: “This day crosses the city three times and overlaps our confirmed entry.” Ask for the smallest repair and a before-and-after explanation. A vague request to make the plan better invites another attractive rewrite that may discard details you wanted to keep.

If a tool can change the trip, separate the planning conversation from the approved action. Inspect the exact preview, identify the right trip and check the returned result. Do not let an example prompt, draft message or hypothetical itinerary become an instruction to book, pay or speak for someone else.

Copyable AI trip prompt

Replace the bracketed fields with your own details. Read the complete text before sending it; copying it here does not send a message or perform an action.

Help with [specific trip task] using only these necessary details: [redacted dates, locations, times, requirements]. Tell me what additional planning field you need and why. Do not request passwords, sign-in codes, payment-card details or unnecessary identity documents. Preview any saved-trip changes before I approve them.

Run a focused refinement

Use a follow-up that names the actual problem, not just “make it better”.

Review your answer about [specific trip task]. The issue is [what failed]. Preserve [confirmed facts and constraints]. Change only [scope], show what changed and why, and list any remaining unknowns. Do not invent sources or approvals. Do not save, send, book or pay.

A connected workflow needs the current account access, permissions and client support described in the Limbo setup guide. If a required tool is unavailable, keep the output as a text draft and do not claim that the trip was saved.

Before you call it settled

• The answer preserves the original trip constraints.
• Important facts have a source or are marked unverified.
• The worked example is not mistaken for a real booking or test.
• Any proposed saved action has been previewed precisely.
• Completion is claimed only when a returned result supports it.

Data minimisation is a practical editing habit: remove the fields that do not help answer the question. Do not claim redaction guarantees privacy or replaces the provider’s actual controls.

Put the decision somewhere useful

Keep the agreed plan and the information needed to act on it together in Limbo. Save the useful outcome of the conversation—not every abandoned option—and make clear what is still awaiting confirmation.

Sources and further reading

• Limbo: current connection setup and supported actions

These references support the workflow and further checks; fictional examples are authored illustrations, not evidence of live prices, bookings or customer outcomes.

Keep planning

• Build an AI trip plan you can actually use
• Group-trip planning guide