I nearly missed a train in Lyon because my itinerary was too perfect.
It was a Tuesday in June, raining sideways, and I had a colour-coded plan that told me to be at a silk workshop at 10:00, a market at 11:30 and a viewpoint at 13:00. What it did not tell me was that the workshop was closed for a staff training day, the market stalls pack up early when it rains, and the viewpoint is just a wet car park in a storm. I stood under a station canopy, phone in one hand, soggy sandwich in the other, and rewrote the whole day in four minutes with one honest prompt: “I am in Lyon, it is raining, I have no car, I like food markets and quiet museums. Give me a loose plan for today with one indoor backup for each stop, and tell me what to skip.”
That rewritten day worked. The backup plan worked because it was built around constraints, not wishes. That is the whole trick with AI travel prompts: the prompt that names your limits beats the prompt that lists your dreams.
I tested every prompt below the same way before writing this up. I fed each one to a chatbot with a real trip in mind — five days in Lyon and the Rhone-Alpes, two adults, one small backpack each, a budget that is comfortable but not careless — and I kept only the prompts that returned something I could actually act on. Where a prompt produced something vague or wrong, I say so. Travel advice that hides its failures is not advice.
A note on how to use these: copy a prompt, then replace the bracketed details with yours. The brackets are doing real work. A prompt without your dates, your pace and your deal-breakers is just a horoscope.
Packing prompts that stop the 2 a.m. suitcase panic
- “I am travelling to [city] for [number] days in [month]. The forecast is [weather]. I will [walk a lot / attend a wedding / work remotely]. Build me a packing list in three groups: wear, carry, and leave at home. Flag anything I can buy there for under 10 euros instead of packing it.”
When I tried this for Lyon in June, it told me to leave the second pair of shoes at home and buy a cheap umbrella on arrival. It was right on both counts. The umbrella cost 7 euros at a tabac near the station, and my feet thanked me for the lighter bag.
- “Here is what I own: [list 10 items]. I need outfits for [number] days that mix and match, with no outfit repeated exactly. Show me the combinations in a table and name the one item I am missing.”
- “I am flying with only hand luggage on [airline]. My bag measures [dimensions]. Tell me what will not fit, what counts as a liquid, and what I should wear on the plane to save space.”
- “Pack for a trip where I will go from a beach in [place] to a smart dinner in [place] on the same day. One bag, no checked luggage, no iron. What fabrics and colours make this possible?”
- “I have [medical need / dietary need / accessibility need]. What should I pack, what should I declare at the airport, and what should I translate into [language] on a small card?”
- “Make me a packing checklist for travelling with a toddler in Europe: nappies, snacks, entertainment and documents, split into hand luggage and checked bag. Add one thing parents always forget.”
- “I am travelling light for two weeks across [countries]. Give me a capsule wardrobe of 12 items, and tell me which two items do the most work.”
- “Check my packing list for mistakes: [paste your list]. What is duplicated, what is missing for [weather], and what will get confiscated at security?”
The honest rule I now follow: if the AI suggests more than 15 clothing items for a week, it is packing for a photoshoot, not a trip. Cut a third and you will still be overdressed.
Itinerary prompts that leave room for real life
- “Plan [number] days in [city] for a first-time visitor who hates queues and loves [food / architecture / nature]. Maximum two big sights per day, one long lunch, and one evening with nothing planned. Include walking times between stops.”
- “I arrive in [city] at [time] and leave at [time]. Build an itinerary around that, not around a full day. Include where to store luggage and what to do first if I am jet-lagged.”
- “Re-plan my day: it is raining in [city], [attraction] is closed, and I am with [kids / elderly parents]. Give me a new plan with indoor alternatives and shorter walking distances.”
This is the prompt that saved Lyon. It replaced the viewpoint with a covered food hall, and the closed workshop with a silk museum that was actually open. The difference between the two plans was one sentence: “tell me what to skip.”
- “I want to see [city] like a local, not a checklist. Give me one neighbourhood to wander with no agenda, three places locals actually eat, and one tourist trap to skip — with a reason.”
- “Compare two options for me: a day trip to [place A] versus a slow day in [city]. Include travel time, cost, and what I would miss in each case. Then recommend one and defend it.”
- “Build a slow-travel itinerary: [number] days, maximum one city change every three days, train travel only. Show me the train times and the one journey worth paying extra for.”
- “I have one free afternoon in [city] and 40 euros. What is the best use of that time and money? Be specific: name places, not vibes.”
- “Turn this rough plan into a timed itinerary with buffers: [paste your rough plan]. Add 30 minutes of slack after every museum and tell me where the day is too tight.”
- “Plan a trip around a single interest: [wine / medieval history / street art / cycling] in [region]. One base town, three day trips, no rental car.”
- “My flight lands late and my hotel check-in is the next morning. What do I do between midnight and 9 a.m. in [city] safely and cheaply?”
Language prompts that get you further than “hello”
- “Teach me 20 survival phrases in [language] for [country]: greetings, directions, ordering food, asking for help, and apologising politely. Include pronunciation in plain English, not phonetic symbols I cannot read.”
When I tested this for French, the chatbot gave me “Je voudrais…” for ordering and, crucially, “L’addition, s’il vous plait” for the bill. The second phrase saved me twice in one weekend. The first phrase I already knew. The useful prompt is the one that teaches you the second sentence, not the first.
- “Write me a polite message in [language] to a hotel: my flight is delayed, I will arrive at [time], please keep my room. Keep it under 40 words and formal.”
- “Translate this, but make it sound like a local wrote it, not a textbook: [your sentence]. Then explain the one word you changed and why.”
- “Role-play with me: I am ordering dinner in [language] in [city]. You are the waiter. Correct my mistakes gently and do not switch to English unless I ask.”
- “What are five things tourists say in [language] that sound rude without meaning to? Give me the polite version of each.”
- “I need to explain a food allergy in [language]: [allergy]. Give me one short sentence for a restaurant, one for a pharmacy, and one written version I can show on my phone.”
- “Teach me how to ask for directions in [language] and understand the answer: numbers, left and right, and the words for ‘near’, ‘far’ and ‘closed today’.”
Money, safety and the boring prompts that matter most
- “I am in [country] for [number] days with a budget of [amount]. Split it across accommodation, food, transport and activities, and tell me where Europeans actually save money that tourists do not.”
- “What scams target tourists in [city] right now, and what is the exact sentence I should use to walk away politely? No scare stories — only specific, current tactics.”
- “My phone was stolen in [city]. Give me a step-by-step recovery plan: what to block first, where to report it, and how to get home without my cards or maps.”
- “Compare travel insurance for a trip to [countries]: what does a basic policy actually exclude? List the three exclusions that surprise people most.”
- “Write a one-page trip brief for my family back home: where I am staying, my travel dates, emergency numbers for [country], and what to do if they cannot reach me. Keep it factual and calm.”
Three prompts that flopped, and why
Not every prompt in this list earned its place on the first try. Three failed in ways worth knowing.
First, “Plan the perfect trip to Europe” returned a 14-day sprint through seven cities that would have cost a fortune and left no time to eat sitting down. The fix was constraints: dates, budget, pace, and one interest. Without constraints, the AI plans for everyone and therefore for no one.
Second, “Translate this menu for me” failed on a handwritten menu in a Lyon bouchon because I photographed it at an angle in bad light. The AI guessed at words that were not there. The fix was boring and human: I typed the three dishes I could read and asked about those, and the waiter translated the rest with a smile.
Third, “Find hidden gems in [city]” returned the same five places every travel blog lists. Hidden gems are hidden precisely because they are not in the training data’s greatest hits. The prompt that actually worked was number 12 above: one neighbourhood, no agenda, and a willingness to get a little bit lost.
My stance, plainly stated
Here is an opinion you are free to disagree with: an AI-planned trip should never be more than 70 percent planned. The best travel day I had in Lyon was the unplanned one — the rainy backup day, the food hall, the long lunch that stretched to three hours because nobody was in a hurry. If your itinerary has no empty hours, it has no room for the trip itself. Use these prompts to remove the boring work — the packing lists, the translations, the backup plans — and then guard the empty hours fiercely. That is where the actual holiday happens.
If you want a fuller walkthrough of planning a whole trip with a chatbot, our step-by-step guide to planning a trip with ChatGPT pairs well with the prompts above, and if your travels take you to Germany in December, the Christmas market trip guide shows how the same loose-planning method works when the weather turns.