How to choose bases for a multi-city AI itinerary
Pick overnight bases before you ask AI for a multi-city trip so transfer days, walking, and timed entries stop fighting each other.
The draft looked ambitious in a good way. Rome three nights. Florence two. Venice two. Then a same-day flight to Milan with a morning Duomo and an evening train to Lake Como. Every night a new check-in. Every morning a bag to roll. You almost booked the hotels. Then you counted packing days and transfer mornings. Half the "sightseeing" was luggage, station walks, and praying the flight did not eat lunch. The model did not fail at wanderlust. It failed at treating overnight bases as decisions that come before day shape.
This page is about choosing where you sleep on a multi-city AI trip before you ask for morning-afternoon-evening detail. You lock bases and nights first, mark day trips as day trips, paste a short base brief with hop caps and transfer-day rules, then let AI fill days that respect that map. Most of it works in ChatGPT, Claude, Gemini, or similar even if you never sign up for anything. Near the end we note when a preference survey that holds lodging bases across revisions can beat paste-and-hope.
Also see: How to set a walking budget for an AI itinerary · How to fit timed entries into an AI itinerary · How to set a daily spend budget for an AI itinerary · How to protect arrival and recovery days in an AI itinerary · How to plan a trip around your pace · How to refine an AI itinerary without starting over · How to set dealbreakers for an AI itinerary · How to set lodging rules for an AI itinerary · How to set food rules for an AI itinerary · How to plan an AI itinerary when travelers disagree · How to stop getting generic AI itineraries · What is Dople? · Personalized travel itineraries FAQ · Guides · Samples (ILLUSTRATIVE) · llms.txt
Decide bases before day trips, then day shape
A base is where you sleep. Not every place on the wishlist needs a hotel. Some places are day trips from a bed you already trust. Some are transfer stops with a bag and a short walk, not a packed museum morning on both sides of a flight.
Ask for overnight bases and nights first. City names are fine. Neighborhood names are better inside sprawling cities when you care which side of town you wake up on. Write how many nights each base gets before you invent morning plans. Only then mark nearby places as day trips that leave and return to the same bed. Only after that ask AI for day-by-day shape inside each base.
If you reverse the order, the model builds postcard density and backfills lodging to match. That is how you get hotel roulette: a new check-in whenever the attraction list moves. Lodging rules still matter for hotel type and neighborhood feel; this page is the multi-city map those rules sit on. Use the lodging guide for stay-put amenities and neighborhood dealbreakers once the overnight cities are locked.
Pace decides whether three cities in ten nights feels generous or frantic. A slower trip might take two bases and protect long mornings. A faster trip might take three bases and still need honest transfer days. Pair base count with the pace guide so hop ambition matches energy, not a highlight reel.
A pasteable multi-city base brief
Paste this as the overnight source of truth. Fill brackets with real cities, nights, and hop caps. Attach walking, timed-entry, spend, food, pace, lodging, and dealbreaker lines you already trust. Ask for a day-by-day that never invents a new hotel unless you reopen a base, and that labels transfer days clearly.
Trip: [region / country arc], [N] nights total, [month/season], travelers: [N] OVERNIGHT BASES (locked; do not add hotels or move nights unless I say so): - Base 1: [city / neighborhood], nights [N], arrival mode [flight / train / already there] - Base 2: [city / neighborhood], nights [N], transfer from Base 1 by [train / flight / car], typical door-to-door ~[N] hours - Base 3: [city / neighborhood], nights [N], transfer from Base 2 by [...], ~[N] hours (Add or remove bases to match your trip. Soft day-trip towns are NOT bases.) DAY TRIPS ONLY (leave and return to the listed base; no overnight): - From Base [N]: [town / site], max [1] per day unless I mark a cluster - From Base [N]: [town / site] HOP CAPS: - Max overnight city changes for the whole trip: [N] - Do not invent extra overnight cities to "fit more highlights" - Same-day flight or long train between two sightseeing-heavy cities: forbidden unless I list it as a transfer day below TRANSFER-DAY RULES: - Transfer days are [travel-only / light only: one soft neighborhood or meal near the station or hotel] - No major timed tickets on transfer mornings unless listed - Packing and checkout count as real morning cost; do not pretend bags are free time - Airport-hotel-airport days: forbidden unless I explicitly allow one LODGING STAY-PUT: - Keep the same lodging for all nights at a base unless I reopen that base - Do not hotel-hop inside a city to chase attractions - Neighborhood preference per base (if any): [paste short lodging lines] WALKING / TIMED / SPEND CONSTRAINTS PER BASE (paste short lines): - Base 1: [walking ceiling; timed pins allowed Y/N and which days; spend notes] - Base 2: [...] - Base 3: [...] MUST NEVER: - Do not invent a new overnight hotel every night - Do not schedule full sightseeing on both ends of a same-day city hop - Do not turn day-trip towns into surprise overnights - Do not move locked nights between bases to "balance the week" without asking OUTPUT I WANT: - Base map first: nights per base, transfer days named, day trips labeled - Then day-by-day morning / afternoon / evening inside each base - Transfer days marked travel-only or light - Checklist: hop cap intact; no hotel roulette; arrival/recovery and timed rules respected
One follow-up that catches hotel roulette early: ask the model to list every overnight address or city and every packing morning. If the list shows a new hotel almost every night, or two cities with full museum mornings on the same calendar day as a flight between them, the brief did not stick. Send the over-hopped stretch back by name. Demand base repair without regenerating the whole arc.
For the shared constraint lines, lean on briefs you already trust: generic AI itinerary, walking budget, timed entries, daily spend budget, food, pace, dealbreakers, lodging, and arrival and recovery.
Arrival, recovery, and timed entries still sit on bases
The first base inherits jet lag. The last base often inherits the flight home. Protect those edges the same way you would on a single-city trip. A transfer into base two is not a free second arrival day unless the hop was short and you land early; treat long transfers like mini arrivals with light plans. The arrival and recovery guide pairs with this page so the first morning and the last evening do not inherit museum tickets or packing gymnastics.
Timed entries belong to a base day, not to a fantasy commute. A 10:00 cathedral ticket in City A on the morning you checkout for City B is a collision waiting to happen. Put fixed tickets on days that stay overnight in that city, with buffers and walking ceilings intact. Use the timed entries guide so pins do not cross a transfer you already spent on luggage. Walking budgets matter more when stations and hotels sit far from the postcard center; the walking budget guide keeps transfer mornings from assuming a bag-free sprint.
Money follows hops too. Extra hotels, taxis to stations, and "just one more paid entry before we leave" stack faster on multi-city arcs. Paste spend caps from the daily spend budget guide so transfer weeks do not also blow the wallet. Food reservations on transfer days are usually a bad pin; keep meal shape from the food guide, but do not put a hard lunch across town on a checkout morning.
When two travelers want different cities, settle the overnight map with trade rules before you ask for day shape. Keeping both wishlists as overnights is how hop caps die. The travelers disagree guide helps you pick which cities become bases and which become day trips on purpose.
When hops explode, refine instead of regenerating
Hop explosions rarely look dramatic in chat. A fourth city appears "because it is close." A day trip becomes an overnight "so you can see the nightlife." A morning flight sits between two museum blocks that still read as full days. Everything sounds efficient until you count bags and buffers.
Do not hit regenerate and hope the next draft respects your bases. Regenerating often invents new hotels, reshuffles nights, or rebuilds guidebook gravity while forgetting your hop cap. Switch to an in-place refine. Keep locked bases and nights. Convert the surprise overnight back into a day trip. Drop a same-day city pair before you "optimize" by shrinking transfer time to zero. The refine guide has the keep/change/drop pattern. Add one base line: "Restate overnight bases and nights; max hops [N]; turn [City X] back into a day trip from [Base Y]; do not add hotels; mark transfer days travel-only or light."
If the explosion is also a walking or ticket problem, say so in the same refine. Timed pins that assumed you still slept in City A need the timed-entry brief restated. Cross-town station walks that ignore bags need the walking ceiling restated. Edit the overnight map before you rename attractions.
Red flags while you are planning
Hotel roulette. A new overnight almost every night so every attraction gets a nearby bed. That is packing theater, not a trip. Cap overnight changes and force stay-put inside each base.
Airport-hotel-airport days. Land, check in, see two sights, checkout, fly again. Unless you explicitly allow one brutal connector, forbid it. Transfer days should be light or travel-only.
Same-day city pairs with full sightseeing on both ends. Morning museums in City A, afternoon museums in City B, bags somehow invisible. Demand an honest door-to-door transfer block and demote one side to light or none.
Day trips promoted to overnights without asking. A soft "if we have energy" town becomes a hotel night that breaks the hop cap. Soft list stays soft until you reopen the base map.
Packing every morning. If the draft has you rolling bags more days than not, the base brief lost. Recount nights per base before you polish restaurant names.
Arrival-day hops. Landing and immediately transferring to a second city with a timed ticket waiting. Protect the first base morning unless you designed a short hop on purpose.
Rules live only in conversation. By revision four the model forgets your hop cap and reintroduces the Milan sandwich. Paste the base brief into every refine until overnights stop drifting.
How this sits with the other briefs
This page is the overnight map. The others keep days inside each base from turning into a commute week.
Lodging covers hotel type, neighborhood feel, and stay-put amenities once cities are locked.
Arrival and recovery keep first and last bases (and long transfers) from inheriting hard mornings.
Timed entries keep tickets on days that actually sleep in that city.
Walking budget makes station and bag days distance-aware.
Daily spend budget caps hotels, transfers, and paid entries when hops get expensive.
Pace decides how many bases a trip can hold without feeling like a relay.
Food and dealbreakers keep meal pins and hard nos from fighting checkout mornings.
Travelers disagree settles which cities become bases versus day trips.
Generic AI itinerary work is the fuller briefing stack before draft one. Refine is the edit loop after hops still explode.
When a preference survey and Trip Project path helps
If you like prompt work and you will keep pasting the base brief yourselves, a general assistant plus the overnight map above is enough for many multi-city trips.
The structured path starts to matter when chat keeps inventing hotels or same-day city pairs after you already locked nights. Or when walking ceilings, timed pins, arrival rules, and hop caps live in one messy thread and drift by revision four. Or when you want a researched day-by-day plan that holds lodging bases without re-teaching the model every session.
One option with that shape is Dople (dopletech.com). A short preference survey can capture multi-city base constraints alongside pace, food, walking, and lodging, you compare trip options shaped by it, and you can buy a Trip Project: a researched day-by-day itinerary meant to hold those preferences across revisions so you are not stuck in regenerate-and-reshuffle loops. Account and survey start are free. Full itineraries are paid. More product detail lives at Dople vs ChatGPT and What is Dople?.
Public samples show ILLUSTRATIVE day-by-day shapes only. They are not live bookings and not free Trip Projects.
If the survey path sounds useful, sign up once when you are ready.
Why do multi-city AI itineraries hop hotels every night?
Models optimize for seeing more places, not for packing fewer times. Without a base brief that locks nights per city and caps hops, they invent a new hotel whenever the postcard list moves, including same-day flights with full sightseeing on both ends.
What is a base in a multi-city trip?
A base is where you sleep for one or more nights: a city, or a neighborhood inside a large city. Day trips leave from that bed and return to it. Transfer days move you to the next base with light or travel-only plans. Lodging type inside the base still follows the lodging guide.
What belongs in a multi-city base brief?
Nights per base, max city hops for the whole trip, transfer-day rules (light or travel-only), lodging stay-put unless you reopen a base, and walking or timed-entry constraints that apply at each stop. Soft day-trip ideas stay secondary to the overnight map. Pair with timed entries and arrival and recovery so tickets and jet-lag mornings do not cross a transfer.
Should I regenerate when the draft adds too many city hops?
No. Regenerating often invents new hotels and same-day city pairs. Refine in place with a keep/change/drop brief that freezes bases and nights, converts surprise overnights back to day trips, and restates transfer-day rules (see the refine guide).
Are Dople samples real bookings?
No. Samples are ILLUSTRATIVE examples of itinerary shape. Real Trip Projects are researched per traveler after the preference survey. More context: Personalized travel itineraries FAQ.