AI Trip Planners: What They Get Right, the Prompt That Works, and What to Verify Before You Pay

A chatbot will write a seven-day itinerary for Lisbon in under a minute, and the first read looks like an hour of work. Some of it is. Some of it is a museum that closed in 2023 and a restaurant that needs a booking three weeks out.

This post covers what the AI planners are good at, the prompt that gets a plan shaped like your trip instead of a tourist checklist, and the short list of things I would verify before spending money on any of it.

What the planners do better than a spreadsheet

Clustering by neighbourhood. The plan puts the museum, lunch and the afternoon walk within ten minutes of each other instead of sending you across the city three times. This is the thing people get wrong by hand, because they pick favourites first and look at the map afterwards.

Realistic days. A decent plan accounts for travel time between stops, opening hours, how long a visit takes, and a break in the afternoon. It will not put a 16:30 museum visit after a 45-minute tram ride to a museum that closes at 17:00, as long as its data is current. That last clause is the subject of the verification section.

Budget as an input. Tell it 150 euros a day for two and the hotels, restaurants and activities change to match. The same city produces a very different week at 60 and at 400.

Your preferences, not the average tourist's. "Street food, no museums, live music, kid-friendly mornings" produces a different plan from the default. The default is the top ten from every guidebook, and you already know those.

A plan B for rain. Ask for it up front. "If day three is wet, swap the beach for the covered market and an indoor food tour" is cheaper to get now than to work out at 08:00 in a hotel lobby.

The tools, and what each one adds to a chat

ChatGPT, Claude or Gemini is where I would start every time. Describe the trip, get a draft, then push back: fewer stops on day two, swap the hike for something flat, cheaper lunches. The plan is only as good as the follow-up questions, and nothing else on this list iterates as well.

Wanderlog puts the itinerary on a map, lets you drag stops between days, and shares the plan with the people travelling with you. It has its own AI generator, but the map and the sharing are the reason to use it. Paste the chat draft in and rearrange it there.

TripIt is a different job. Forward your flight, hotel and car confirmations and it builds the timeline for you. It recommends nothing. For a trip with four flights and three hotels it is the one tool I would not skip.

Layla is a chat planner with booking links attached to the suggestions. Convenient when you want to go from idea to reservation in one place, less flexible than a general chatbot when the plan needs real reshaping.

Google Maps and Search now surface AI-built trip suggestions and saved lists. Useful for discovery and for the day itself, when you want directions and opening hours in one app. The features change month to month, so I would not build the plan there.

The workflow that works: draft in a chatbot, rearrange and share in Wanderlog or a shared Maps list, get around on the day with Maps. While you are pricing the draft, the currency converter turns "3,500 yen for lunch" into a number you can budget with.

The prompt that gets a plan shaped like your trip

The draft is only as specific as the prompt. Put all of this in one message:

• Destination and exact dates. "Tokyo, 5 to 12 October 2026" beats "Japan in autumn". The dates decide the weekday closures, the weather and the festivals.
• Who is going. Two adults and a toddler, a solo traveller, six friends in their thirties. Each of those changes the restaurant list.
• Budget as a number. "Mid-range" means nothing. "250 euros a day for two, including the hotel" does.
• What you like and what you want to avoid. Both halves. "Avoid: tour groups, nightclubs, shopping malls" removes a third of the default plan.
• Pace. Three planned things a day, or a full schedule, or alternating busy and empty days.
• How you move. Rental car, public transport only, or walking distance from the hotel.
• Dietary restrictions, if any, so the lunch suggestions are usable.

A prompt that has all of it: "Plan five days in Barcelona in July for a couple in their thirties. 250 euros a day in total. We like food, architecture and beaches, and we dislike large tour groups. Relaxed pace, no more than three planned activities a day. Public transport only. Give lunch and dinner suggestions for every day, with the neighbourhood and the price level."

Run the dates through the date calculator first for the exact number of nights, and look at the weekdays. Landing on a Sunday and leaving on a Monday decides which market days you hit and which museums are closed on your only free morning.

What to verify before you pay for anything

The itinerary is a set of claims. I treat it the way I treat a status report: plausible until checked, and checked in the order of what it costs to be wrong.

1. Opening hours and closing days. The model's data has a date, and the plan does not tell you what it is. Museums, churches and government buildings close on specific weekdays. Check the venue's own site for every fixed appointment in the plan.
2. Season. A trail that is closed in winter, a beach that is 14 degrees in March, a boat that does not run in the off-season. The plan does not know the month unless you told it, and even then it guesses.
3. Bookings. Popular restaurants and timed-entry attractions need a reservation, two to four weeks out in high season. The plan may say so. It does not book.
4. Local holidays and events. A national holiday or a festival makes the centre unwalkable or closes everything. One search for the destination plus your dates covers this.
5. Travel times. The plan's estimates are approximations. Put the two addresses in Maps for anything where the timing is tight.
6. Safety and scams. Models occasionally recommend an area or an activity that recent travellers warn about. Ten minutes in a current forum thread for the destination is enough to catch it.
7. Prices. Any number in the plan comes from the training data. Check the venue or the booking site before it goes in the budget.

A plan that is 70 percent right in ten minutes, plus half an hour of checking, beats eight hours of research. The checking is the part you cannot skip.

FAQ

Can the AI book flights and hotels?

Mostly not. The chatbots recommend, you book on Google Flights, Booking.com or the airline. Layla and a few others attach booking links. Agents that book end to end exist in demos and early products, and I would not hand one my card yet.

How far ahead should I plan?

Two to three months for an international trip, which leaves time for visas, reservations and a second pass on the plan. Two to four weeks for a domestic trip or a long weekend. Last minute works too, with fewer options for the places everyone wants. A countdown timer set to the booking deadline for the one restaurant or museum that sells out is more use than one set to the departure date.

Does it work for road trips?

Better than for cities, because ordering the stops is the whole problem and the model is good at it. Say you are driving, give a maximum driving time per day, and ask for stops along the route rather than only at the destinations.

How do I share the plan with the people I am travelling with?

Ask the chatbot to output the itinerary as a table with times, addresses and links, then paste it into Wanderlog, a shared spreadsheet or a shared Google Maps list. Everyone can edit it there, and the day-of version lives on the phone that has Maps open anyway.