September 19, 2026 · Limbo Crew
How to use AI to build a travel plan you can actually use
Build a useful AI travel plan with a complete brief, copyable prompts, a worked example and practical refinements before saving to Limbo.
A useful travel plan is not the longest answer an AI assistant can write. It is a sequence of decisions you can understand: what matters, what fits, what is uncertain and what you are ready to save.
Start with a brief, ask for a first plan, challenge it with one concrete change and verify the important dependencies. Only then move the approved version into Limbo. ChatGPT or Claude can help with the thinking; your constraints and checks make the result useful.
Use ChatGPT or Claude for the conversation, then follow the same review process. This guide is about making a usable plan, not claiming that one provider always produces a better holiday.
The method: brief → first plan → specific refinement → verify → approve and save.
1. Give the assistant a brief it can make decisions with
Start with three useful inputs:
- Destination, dates, arrival and departure
- Travellers, budget and real constraints
- Must-dos, trade-offs and the pace you want
Expand those bullets into constraints, preferences and choices. A constraint is something the plan must respect: a departure time, a child’s rest period or a hard spending ceiling. A preference is something you are willing to trade, such as a second museum in exchange for a slower afternoon. An unresolved choice should become a question, not a hidden assumption.
For this example, the brief is Lisbon, 8–12 December 2026, two adults aged 31 and 35. The travel boundaries are: arrive Lisbon Airport at 10:00 on 8 December; depart at 18:00 on 12 December. The requirements are one hotel near useful transport; food, architecture and relaxed neighbourhood time; avoid long steep walks; one major paid attraction; slow mornings; public transport where practical; rain alternative.
If the first answer skips an important detail, stop there and supply it. A route built on the wrong airport, room arrangement or traveller count is not improved by adding more detail later.
2. Copy the complete prompt, not just the destination
Replace every bracketed field below. Keep the questions, budget checks and approval step; those are what turn a request for inspiration into a planning workflow.
Plan [location] from [start date] to [end date] for [number of travellers + ages], arriving [time / airport] and leaving [time / airport]. Ask any missing questions first.
Budget: [amount + currency] per person, [including / excluding flights]. We love [interests / food]. Must-dos: [list]. Avoid: [deal-breakers]. Allow for [dietary / accessibility needs]. Our pace is [relaxed / balanced / busy], transport [preference], and stay [style / area].
Build a day-by-day itinerary in local time. Group nearby stops to limit backtracking. Include realistic travel times, meal breaks, estimated costs per person and a total budget. Leave time to wander; do not fill every hour. Add a rainy-day alternative.
Link sources for opening hours, current prices and bookings. Flag uncertain details, booking deadlines and anything to verify before paying. Suggest trade-offs if our budget or must-dos conflict.
Help me refine the plan. Once Limbo is connected, preview the itinerary as [new trip name / existing trip name]. Ask me to confirm the correct trip. Do not save yet: wait for my explicit approval of the final preview, then save that exact plan to Limbo and return its link.
Connector setup instructions: https://www.getlimbo.app/mcp
Do not book or pay.
3. See a fully filled-in version
Fictional worked brief. The dates, people, flight times and spending limits below are authored examples. They are not a real booking, a live price quote or a transcript from an AI account. Adapt them before using the prompt.
Plan Lisbon for 8–12 December 2026 for two adults aged 31 and 35. We arrive Lisbon Airport at 10:00 on 8 December; depart at 18:00 on 12 December. Treat all times as local to the place concerned. These are fictional planning inputs, not confirmed bookings. Ask up to five important missing questions before producing a final itinerary; do not quietly invent our preferences.
Our planning ceiling is AUD 1,000 per person, AUD 2,000 for 2 travellers, excluding international flights, travel insurance and personal shopping. For a family, the per-person amount is a planning average, not a claim that adult and child prices are identical. Requirements: one hotel near useful transport; food, architecture and relaxed neighbourhood time; avoid long steep walks; one major paid attraction; slow mornings; public transport where practical; rain alternative.
Compare two practical versions of the same trip. Protect the stated pace and explain each trade-off rather than adding more attractions. Apply Christmas or New Year checks only if they overlap our dates; do not add a celebration outside our stay.
Build local-time day-by-day plans with one main focus, nearby stops, meals, realistic travel allowances, free time and a bad-weather alternative. For each fixed booking, show its source, date, inclusion and cancellation questions. Distinguish an illustrative estimate from a verified live quote. Itemise accommodation, food, transport and activities without double-counting package inclusions. Reserve AUD 120 per person as contingency. If current quotes exceed the ceiling, show what must change rather than forcing the numbers to fit.
Use official sources for opening times, ticket rules and holiday changes. Label anything unconfirmed. Explain trade-offs and invite a refinement round. Do not book or pay.
Once Limbo is connected, preview a new trip named “Lisbon — December planning draft”. Ask me to confirm the correct trip and the exact final version. Wait for my explicit approval, then save that exact plan to Limbo and return its link. If anything fails, report what did and did not save; do not claim completion without a returned result.
A good response should make the day structure easy to scan and expose what it cannot yet confirm. If browsing is unavailable, ask for a planning skeleton and a verification list—not invented sources or confident opening times.
4. What a useful first result could look like
Simulated outcome—not real AI output. This is an editorial example of a planning skeleton for the brief above. Every visit and transfer remains subject to verification. The table shows the flow; it does not assert that a ticket, room or transport service is available.
| Date / block | Main plan | Space around it | What still needs checking |
|---|---|---|---|
| 8 Dec | Arrive, store bags and settle in | One nearby meal and optional short walk | Check hotel access and the airport transfer. |
| 9 Dec | One compact central-neighbourhood day | Food and unplanned café time | Check the walking route rather than distance alone. |
| 10 Dec | A Belém-area outing | One paid attraction if agreed; a relaxed afternoon | Verify opening and entry before assigning the date. |
| 11 Dec | Flexible museum or food-led day | Use it as the rain or energy-level swap day | Avoid prepaying for both the original and backup. |
| 12 Dec | Breakfast, checkout and airport transfer | No distant final stop | Confirm the real departure route. |
Before turning the skeleton into a timed plan, ask for separate blocks for the main activity, meals, travel and downtime. If a journey is not checked, label the duration as a provisional allowance. Do not let an attractive-looking timetable disguise missing transport information.
The useful test is simple: if the main activity takes longer or the weather changes, can you still eat, rest and reach the next fixed commitment? If the answer is no, remove something before adding precision.
Make the budget auditable
The figures below are invented allocations within the fictional traveller’s ceiling, not market estimates or a claim this holiday can be bought for this amount. They include contingency but exclude international flights, travel insurance and personal shopping. For mixed-age groups, the per-person column is an average allocation; real adult and child prices must be priced separately.
| Category | Per-person planning allocation | Whole-party allocation |
|---|---|---|
| Accommodation allocation | AUD 360 | AUD 720 |
| Food allocation | AUD 240 | AUD 480 |
| Local / intercity transport allocation | AUD 100 | AUD 200 |
| Activities / admission allocation | AUD 180 | AUD 360 |
| Protected contingency | AUD 120 | AUD 240 |
| Total ceiling | AUD 1,000 | AUD 2,000 |
Ask for a second table of dated supplier quotes when you are ready to research real costs. If a package includes meals, transport or admissions, remove those items from the other rows before comparing totals. Keep the contingency available; do not spend it simply to make the schedule look fuller. If the actual costs exceed the ceiling, revise the scope or budget openly.
5. Refine one real problem instead of saying “make it better”
Simulated first-draft problem: The first answer includes a day trip, three paid attractions and a steep walking route despite a relaxed brief.
Give the assistant an explicit replacement instruction:
Keep one Lisbon base, one major paid attraction and slow mornings. Remove the day trip. Replace the steep walking route with a transport-assisted option and tell me what we are trading away. Show a rain swap that does not duplicate tickets.
What the improved version changes: The improved version is shorter, more specific and easier to trust. It explains the removed items, protects the traveller’s priorities and leaves the uncertain bookings visibly unresolved.
Keep these three refinement checks in view:
- Group nearby stops
- Keep time to wander
- Verify hours, prices and booking conditions
Ask the assistant to identify what it removed, what it kept and whether the change affects cost or bookings. That small explanation is useful: it prevents a seemingly improved plan from silently abandoning the one experience you cared about most.
A second refinement can test a disruption. Ask: “If our main outdoor block cannot happen, replace that block without adding another hotel move, exceeding the budget or duplicating prepaid tickets. Show any remaining uncertainty.” A genuine backup should fit the same constraints as the original day.
6. Check the details that can break the plan
For this guide, the priority checks are: Live prices and availability; exact opening dates; travel times and accessibility; budget arithmetic; backup costs; the correct Limbo trip and the final version to save.
Use three labels when reviewing the answer:
- Confirmed for our date: supported by an official source or an actual supplier confirmation, with the date checked.
- Planning assumption: useful for comparison, but not a verified service, booking or price.
- Decision still needed: something the travellers must choose, such as a hotel base, paid upgrade or shared cost.
Do not ask the AI to convert the second and third labels into the first just to make the itinerary look finished. A visible unknown is more helpful than false certainty.
Official places to continue checking
These are planning references, not endorsements of a specific supplier or evidence of available inventory. General tourism pages do not verify every venue in the simulated itinerary. Open the relevant operator page for the exact ticket, transport leg or accommodation you choose. Photos provide destination inspiration; they do not establish seasonal conditions or form a mandatory route.
7. Review the final version, then save it to Limbo
Keep the final review separate from the save action. First confirm the destination, dates, travellers and intended trip. Then read the itinerary that will actually be saved—not an earlier answer from the conversation. For a group, a generated plan is still a proposal; it is not evidence that everyone has agreed.
Follow the current Limbo connection instructions before requesting a save. The current connection requirements summarise account and access requirements. If the connection is not available, retain the text draft rather than claiming anything was saved.
Use this final handoff prompt:
Once Limbo is connected, preview the revised Lisbon itinerary for 8–12 December 2026 as a new trip named “Lisbon — December planning draft”. Show the dates, travellers, day-by-day entries, local times, costs and unresolved booking checks. Ask me to confirm the trip and this exact version. Do not save until I explicitly approve. After approval, save only the approved version and return the trip link. Do not book, pay, send invitations or make additional changes.
After a successful save, open the returned trip link and compare the saved dates, day structure and any unresolved items against the preview. If only part of the plan saved, resolve that explicitly before retrying; do not assume another request cannot duplicate entries.
The next step: choose your dates, complete the reusable prompt and work through one refinement before saving. The goal is a plan your travellers can understand and improve together—not a perfect-looking answer on the first attempt.