How to Use Claude as a Travel Planner (and What to Verify Before Booking)

Skill · SKILL.md

The skill behind this guide: Sofia - Trip Planner AI Skill. Plans any trip end to end in Claude, ChatGPT, or any AI chat - $14.99, one payment, yours permanently.

View the Sofia skill →

Claude will draft a full day-by-day itinerary in under two minutes, and the shape of it is usually sound. What it cannot do is tell you today's opening hours, this season's ticket price, or whether an entry rule changed last month - it recalls those from training data and states them in exactly the same confident tone as the name of the capital city. Use it for structure, sequencing and trade-offs. Verify every date, price and rule on the official source before you pay for anything.

That split - reliable on shape, unreliable on anything with a date attached - is the whole craft of planning a trip with an AI. The prompts below are built around it.

What does Claude actually get wrong when it plans a trip?

It rarely invents a city. It gets the shape right and the details stale. A model answers from a snapshot of text collected months or years before you asked, so a museum that moved its closing day from Monday to Tuesday, a ticket that went from €18 to €26, a ferry that switched to the winter timetable - all of it comes back written with the same certainty as a fact that has been true for a thousand years.

Turning on web search helps and does not solve it. The model can land on the operator's own page or on a travel blog from 2019, and it will not always tell you which.

Treat every one of these as unverified until you have checked it yourself:

  • Opening hours, weekly closing days, seasonal shutdowns and public holidays
  • Ticket prices, and whether entry is timed or sells out weeks ahead
  • Visa status, ESTA / ETA / ETIAS requirements, and passport validity rules
  • Border crossing hours and what you are allowed to carry across
  • Ferry, train, bus and internal flight timetables
  • Driving times, toll costs, and mountain pass or road closures
  • Whether a named restaurant, hotel or shop is still trading at all
  • Airline baggage allowances, fare conditions and change rules
  • Health and vaccination entry requirements

Official source means the operator, not an aggregator: the museum's own site, the airline's own fare rules, the destination government's immigration pages. Aggregators go stale too, and they go stale quietly.

How do you make it flag its own guesses?

Ask for it explicitly. A model will separate durable knowledge from time-sensitive claims quite happily - it just never does it unprompted. Run this straight after it hands you a plan.

PROMPT - the verification pass

Go back through the itinerary you just gave me and sort every factual claim into two lists.

STABLE - things that do not change month to month: where a place is, roughly how long a walk takes, what a neighbourhood is like, general seasonality, how the transport system is laid out.

VERIFY - anything with a date, a price, a timetable, an opening hour or a legal rule attached.

For each VERIFY item give me four things: the exact claim, your confidence (high / medium / low), roughly when you believe it was last true, and the specific official page I should check it on - the operator or the government site, not a review aggregator.

Do not flatten this by marking everything medium. If you are recalling a number rather than reading one, say so plainly.

The VERIFY list is your booking checklist. Work down it before you spend money, and the trip stops depending on what a model remembered.

What do you tell it before it plans anything?

Generic AI itineraries read like brochures because the model committed to a structure before it knew the trip. Four days and fourteen days are different animals. So are a toddler and three university friends. So is €60 a night as a hard ceiling versus €60 a night as a warm-up. Give it everything in one message and the first draft is already close.

PROMPT - the trip brief

Plan a [NUMBER]-day trip to [DESTINATION], arriving [DATE], leaving [DATE].

Who is going: [two adults, one 4-year-old]
Budget: [TOTAL] excluding flights - about [AMOUNT] a day for accommodation, food and entry
Base: [one place the whole time / moving every 2-3 nights]
Pace: no more than [NUMBER] ticketed activities per day. We [do / do not] do early starts.
The trip is for: [food / hiking / museums / doing nothing]
Dealbreakers: [long transfers, crowds, anything before 9am]

Build it to these rules:
1. Group each day by area so we are not crossing the city twice.
2. State travel time between consecutive stops and how we get there - walk, metro, taxi.
3. Give a running cost for each day and a total at the end.
4. Name specific places. Never write 'explore the old town'. One line on why each stop earns its slot.
5. Separate what must be booked more than a week ahead from what we can walk into.
6. Where you are unsure a place is still open or a price is current, say so on the line itself instead of guessing.

What does the itinerary look like when it works?

Each day gets a shape instead of a list: one area, a morning, an afternoon, an evening, and travel time counted between them rather than assumed away. A running cost shows where the money goes before you commit. Booking-ahead items sit in their own list. And because it lives inside a conversation, 'day three is too packed' or 'we added Seville' reshapes the week in place rather than restarting it.

Skill · Keep The Whole Trip Organized
Diane - Executive Assistant AI Skill
Diane - Executive Assistant AI Skill
$14.99this skill vs $47,460/yr medianhiring an assistant

Once the itinerary is settled the admin starts: confirmation numbers, check-in times, the VERIFY list above. Diane keeps all of it in one place instead of six screenshots. The comparison figure is the US median annual wage for secretaries and administrative assistants, $47,460 in May 2024, from the Bureau of Labor Statistics Occupational Outlook Handbook.

View Diane →

How do you get trade-offs instead of one plan?

A single itinerary is a recommendation you cannot argue with. Ask for versions and their costs and it becomes something to negotiate.

PROMPT - the trade-off pass

Take the itinerary above and give me three versions of it, same dates, same people.

1. CHEAPER - cut [AMOUNT] from the total. Tell me exactly what I lose and which single cut I am most likely to regret.
2. SLOWER - one fewer activity per day, fewer places, more time in each. Show me what drops off entirely.
3. WET - the same days rebuilt for a forecast of rain all week, indoor-heavy, same budget.

For each version: total cost, the biggest compromise in one line, and who it suits.

Then tell me which one you would take and why, in two sentences. Commit to an answer - do not tell me it depends on my preferences.

Where does it fail?

  • It cannot see availability. It has no idea whether that hotel has a room on your dates or whether the timed-entry slot is gone. Every 'book this' is a task for you, not a confirmation.
  • It cannot price anything live. Flight and hotel figures are estimates recalled from training data, not quotes. Treat them as a planning range, then check.
  • It gets worse the smaller the place. Major cities are dense in training data. A 900-person village is where invented restaurants and imaginary bus routes appear.
  • It will not volunteer that a plan is bad. Ask 'what is wrong with this itinerary?' and it finds real problems it happily left in a moment earlier.
  • It cannot give immigration or legal advice. A visa answer from a chat window is a starting point for research and never the answer.
  • It has no opinion about you it did not get from you. A vague brief produces an average plan, every time.
Skill · Book What Sofia Planned
Marco - Travel Agent AI Skill
Marco - Travel Agent AI Skill
$14.99this skill vs $100-$300 per tripa travel advisor's fee

Marco covers the booking side Sofia does not touch - fare classes, what basic economy actually forbids, booking windows, hotel positioning, and which reservations are worth paying to keep flexible. AAA puts a travel advisor's trip-planning fee at $100 to $300 for a standard getaway, and $300 to $700 or more for multi-destination or international trips.

View Marco →

How do you install the Sofia skill?

Sofia is a SKILL.md file inside a ZIP. In Claude, open Customize → Skills, press the + button, choose Create skill, then Upload a skill, and select the ZIP. Toggle it on and it is live. Skills are available on the Free, Pro, Max, Team and Enterprise plans.

The one real requirement is code execution. If the Skills menu is greyed out, turn it on under Settings → Capabilities, where it is listed as code execution and file creation. On Team and Enterprise plans an owner controls it in Organization settings, so if it is switched off org-wide, that is who to ask. Anthropic documents all of this in its Skills help article.

If you cannot enable code execution, or you are working in ChatGPT or Gemini, the fallback uses the same file: open the ZIP, copy the contents of SKILL.md, and paste it into custom instructions. The behaviour is identical - you just re-paste it when you move tools.

Planning the same trip by car?

Everything above is about designing a trip. The moment the trip becomes a drive, the constraint changes: the day is governed by how many hours you can stand behind the wheel and whether you arrive before dark. That is a different planning problem, covered in the road trip planner guide. For the money, see the travel budget planner guide; for entry rules, the visa and entry planner; for what to bring, the packing list builder; travelling with children, the family travel planner. The full travel skills collection is here.

In summary:

Use Sofia ($14.99) to build the day-by-day itinerary, then run the verification pass and check every dated fact on the operator's own site. Marco ($14.99) handles the booking side, and Diane ($14.99) keeps the confirmations straight. All work in Claude, ChatGPT and any AI chat, no coding, 30-day money-back guarantee.

Skill · SKILL.md · Works with Claude & ChatGPT

Sofia - Trip Planner AI Skill

One file, loaded once. It interviews you first, then builds the trip day by day - budget-aware, distance-aware, and yours to reshape on demand. No subscription.

$14.99
Get the Sofia skill →

KissMySkills is a marketplace of 853 AI skills, 158 prompt packs, 55 agents & free tools for Claude, ChatGPT & any AI chat.

Related Skill Guides

Frequently asked questions

Can Claude book flights or hotels for me?+

No. It has no booking access, cannot see live availability, and cannot see current prices. Any figure it gives you is an estimate recalled from training data. What it produces is the itinerary plus a list of what needs booking and by when — you do the booking on the operator's own site.

How accurate is an AI-generated itinerary?+

The structure is usually sound: sequencing, grouping stops by area, pacing a day, budget shape. The time-sensitive details are not. Opening hours, closing days, ticket prices, visa and entry rules, ferry and train timetables, and whether a restaurant still exists all change constantly, and the model states stale versions with total confidence. Check every one of those on the operator or government site before you pay.

Does turning on web search fix the accuracy problem?+

It helps but does not fix it. The model may pull the operator's own page or a travel blog from several years ago, and it will not always tell you which. Web search narrows the gap; it does not remove the need to confirm dated facts at source.

How do I get Claude to admit what it is unsure about?+

Ask it directly, after it has produced the plan. Tell it to sort every claim into STABLE (things that do not change month to month) and VERIFY (anything with a date, price, timetable or legal rule attached), and for each VERIFY item to state its confidence and name the official page you should check. It will do this reliably when asked and essentially never on its own.

Do I need a paid Claude plan to use the Sofia skill?+

No. Anthropic makes Skills available on the Free, Pro, Max, Team and Enterprise plans. The actual requirement is code execution: upload the ZIP under Customize > Skills and toggle it on, and if the Skills menu is greyed out, enable code execution and file creation under Settings > Capabilities. On Team and Enterprise plans an owner controls it in Organization settings. If you cannot enable it, or you are using ChatGPT or Gemini, paste the contents of SKILL.md into custom instructions instead.

What does a skill do that a good prompt does not?+

It persists. A skill holds your budget, pace, dealbreakers and everything already decided across the whole conversation, so 'make day three cheaper' or 'we added another city' reshapes the existing plan instead of triggering a fresh generic one. A one-off prompt is fine for a single first draft; a multi-week trip that keeps changing is where the difference shows.

~/get-started

Skills that work. No fluff.

Browse every skill, prompt pack, and agent in the store.

Browse all skills →Or start with free skills