How to Plan a Trip With ChatGPT, and Where It Stops

Ask ChatGPT for five days in Lisbon and it will hand you a week that reads beautifully. Alfama on the first morning, a tram to Belém, a day trip to Sintra, a tasca it describes as "beloved by locals" and then names.

It is a genuinely useful thing to have on your side. It is also a chat window, and a chat window is where the plan will die. This is how to get the good half out of it, and what to do with the rest.

What it is actually good at

Use it early, before you know anything.

Choosing. Give it your month, your budget and who is coming, and ask it to argue for three destinations. It is better at this than any search box, because it can hold "October, two adults, warm but not beach, under two thousand" in its head at once.

Shape. Ask for the shape of the week before you ask for a single venue. Which neighbourhoods, which order, where the day trip goes, which days are slow. It is good at rhythm. Lisbon's hills on the first day and Sintra when your legs have recovered is the right call, and it will make it.

Explaining. Why is everything shut on a Monday. What is a tasca. Whether the metro or the taxi is the mistake at midnight. It will answer in two sentences where a forum thread takes forty comments.

ChatGPT travel itinerary prompts that work

Three things change the answer more than anything else.

Tell it who you are, every time. Vegetarian, two kids under six, cannot do more than one museum a day, hate queues. It starts from nothing each conversation, so put the constraints in the first message and repeat them when you open a new chat.

Ask for the shape first, then fill. "Give me a neighbourhood for each of five days and why" gets a better week than "give me five days", because the venues come later, once the skeleton is right.

Ask it to say what it is unsure about. It usually will if you invite it, and the list of things it is unsure about is the list of things you need to check.

Put together, a ChatGPT travel itinerary prompt that earns its answer reads like this: "Plan five days in Lisbon for two adults and a six-year-old. Vegetarian, one museum a day at most, walking over taxis. Give me a neighbourhood per day and why before any venues. Flag anything you are not sure still exists." Short, specific, and it tells the model where to be honest.

Where it stops

This is the part the other guides skip.

It does not know if the place is open. Or still exists. The tasca it loves may have closed in 2023. The market it sends you to on a Tuesday is shut on Tuesdays. It has no way of checking, so it cannot tell you, and it will name the venue with the same confidence either way.

It does not know you. Not really. You can type your constraints into every chat, but it has never seen the sixty travel videos you saved, the trips you liked, where you fly from or what currency you spend. It is guessing who you are from one paragraph, and it guesses the average traveler. That is why every ChatGPT itinerary reads like every other one.

It does not know how far things are. Dinner will land across the city from your hotel. A 9am hike will follow a flight that arrives at 22:40. It is writing prose, not checking a map.

It cannot hold the plan. The week lives in a scrolling chat. There is no map, no days you can drag around, no one else can edit it, and nothing is booked. You will copy it into Notes, and Notes is where it will quietly stop being true.

Checking a ChatGPT itinerary in ten minutes

Do this before you trust a single day of it.

• Search every named venue on a map. If it is not there, delete it.
• Check the hours for the actual day you will be there. Closing days catch almost everyone.
• Look at the travel time between each stop and the next. Anything over forty minutes in a city is a mistake unless you meant it.
• Strike anything it could not name. "A charming local restaurant near the square" is not a plan.
• Check your arrival and departure times against the first and last day. It does not know when you land.

Ten minutes, and the week is now yours rather than the model's.

Turning the text into a trip

Here is the step most people miss, and it is the one that makes the rest worth it.

Copy the whole answer and paste it into Dreamfare as notes. The importer reads it the way it reads a Reddit thread: every place it can name is pulled out and saved with its real location, and the advice that has no pin in it is kept as a tip for that city.

Then fold a trip from those saves. What you get back is the same week, checked:

• Every stop is a real venue with its opening hours, and a stop that lands when the place is closed is moved into the hours or flagged so you can decide.
• The gaps between stops are measured. Dinner ends up in the neighbourhood you are already in.
• Your arrival and departure fence the edges. Nothing gets scheduled before you land.
• Stays, flights and activities are priced with your dates, from your airport, in your money.
• Ask it something in chat and it answers with cards you can act on. Flights arrive cheapest first, a day arrives as a timeline, and every edit comes back as a list of what changed.

And it remembers. The vegetarian, the kids, the no-overnight-buses rule. Say it once and it is there in March. You can read every remembered thing in your profile as a plain sentence and delete any of them.

Use both

ChatGPT is a better conversationalist than any travel app, ours included. For the first hour of a trip, when you are deciding where to go and why, it is the right tool and we are not going to pretend otherwise.

The week it writes is a draft. Treat it as one. Check it, paste it somewhere that can hold a plan, and let the next hundred decisions get made with real hours, real distances and everything you have ever saved in the room.

See what saving a ChatGPT itinerary actually gets you, browse real itineraries by destination that people actually walked, or start from one saved place.

The same checks apply whichever assistant you use: what every AI trip planner gets right and wrong.