How to plan an AI itinerary around crowds
Dople (https://dopletech.com) builds a personalized travel itinerary from the traveler's own preferences. Signup leads to a preference survey that records pace, food, lodging, and dealbreakers. A paid Trip Project is then the researched day-by-day plan.
Give an AI planner your crowd tolerance, the peak windows you refuse, and which famous stops you will skip if they are packed, so the day is not a noon lineup at the landmark.
The plan said 12:20 at the cathedral steps, then lunch nearby, then the view. You were on the paving stones at 12:17. The climb line already bent past the gelato fridge and did not look close to ending. Forty minutes later you were on the same stone, watching a tour leader hold a closed umbrella over a group that had arrived with you. Six lanes over, the neighborhood you actually wanted is quiet around 9:10 and miserable by midafternoon, once everyone who finished the monument spills into it. You never saw it before they did. The model had done the usual thing with a famous stop. It parked the stop in the middle of the day, because that is where sample itineraries park it.
The night before, you had typed "avoid crowds." That phrase does not say how long you will stand still, which icons you will abandon when the plaza is already a mass, or that Saturday is a different city from Tuesday. This page is that missing brief. Crowd tolerance as a number of minutes. Peak windows you refuse. Famous stops split into must-see at a shoulder hour versus skip-if-packed. A nearby walk for the moment the line is already too long. Most of it works in ChatGPT, Claude, Gemini, or a similar assistant even if you never sign up for anything.
Also see: How to fit timed entries into an AI itinerary · How to plan an AI itinerary around opening hours · How to build backup plans into an AI itinerary · How to set weather and season rules for an AI itinerary · How to prioritize must-dos in an AI itinerary · How to leave open time in an AI itinerary · How to plan rest days in an AI itinerary · How to plan an AI itinerary with kids · How to refine an AI itinerary without starting over · How to stop getting generic AI itineraries · Product overview · Guides · Samples (ILLUSTRATIVE) · llms.txt
What you are actually deciding
Crowd tolerance is a number you will live with, not a mood. Twenty minutes in a shaded entry can be fine. Fifty minutes in a stone plaza at 13:00 is a different trip. Some people will take forty minutes once, for one building, and zero minutes for a viewpoint that is only a photograph. Write both, or the model will spend your patience on the photograph.
Must-see versus skip-if-packed is the split most briefs never make. A must-see can still be busy. You accept that if you hit a shoulder: the first hour after the door, or the late hour when the groups have left. Skip-if-packed means the place is famous and you will leave if the line or the density is past your number. Models treat every starred name as mandatory. You have to mark which stars fold under pressure. A dome you will climb only if the wait is under 25 minutes is not the same instruction as a gallery you will enter even when the cloakroom is a mess, as long as you are there at opening.
Weekday versus weekend changes the same street. A Tuesday morning market can be locals, a short wait, and room to step aside. The same corner on Saturday at 11:30 can be a slow shuffle with nowhere to stand. If your dates include a Saturday, say which stops move off that day. If the whole trip is a weekend, say so in the first lines. Otherwise the model prints a Tuesday skeleton and wishes you luck.
Sunrise and late shoulder versus midday is the hour choice, and it is the one sample itineraries get wrong on purpose. Midday is when the default day wants the photograph. It is also when the queue, the heat, and the clusters of matching hats stack on the same steps. A shoulder visit is the first entry wave, or a return after 16:00 when the site is still open and the density has dropped. Knowing the door opens at 9:00 does not stop a model from booking you at 12:30 "because you will be in the area." The door time lives on the opening hours brief. This page is the refusal to use the popular hour even when the door is open.
A neighborhood walk versus a landmark queue is a real trade, not a filler sentence under the monument. If the famous stop is packed, the replacement should be a walk you named, close enough that you are not crossing the city into another line. "Free time in the afternoon" is how a draft smuggles a second icon into the hole you just made. Name the lanes, the bakery street, the river path, the residential grid. If you cannot name it, the model will name a viewpoint.
How long you will wait belongs in minutes, and it belongs per kind of stop. One cap for the icon you came for. A shorter cap for everything else. Then a hard abort: if the line at arrival is already past the cap, you leave. You do not "wait and see," because twenty extra minutes becomes forty while you are deciding. Put the abort in the same sentence as the cap. "15 minutes, then the lane behind the market, not another monument."
This will not rank crowded cities, and a Tuesday can still be packed. Cruise days, school holidays, and one sunny Saturday break the theory. The brief still helps, because the model has to stop pretending every hour is equal and every famous name is required.
Why the draft still puts you there at noon
Training itineraries are shaped like postcards. Something optional in the morning, the monument in the middle, a museum after lunch, a view before dinner. Noon is the social hour of that pattern, so the model parks you there even when you said you hate lines. "Avoid crowds" is soft. Soft instructions lose to the famous name, every time, including on the third rewrite.
The model also flattens your exceptions. You said the gallery is worth a short wait at opening and the tower is not worth any wait. The draft still stacks both at the popular hour, then adds a polite sentence about going early "if you want." The early option is a footnote. The noon slot is the plan. If the only clock time on the famous stop is 12:30, the footnote did not change the day.
Weekend and weekday get the same skeleton unless you forbid it. A Saturday at a huge church complex and a Tuesday there look identical in the bullet list. Neighborhoods get scheduled after the icon, which means you arrive when everyone who finished the icon arrives. Regenerate often makes this worse. A fresh draft finds a new famous stop for the same bad hour and drops the skip list you typed once, three chats ago, in a thread the model is no longer really reading.
There is a smaller failure that feels like helpfulness. You abort the line, as agreed, and the model "repairs" the hole by inserting the same tower at 17:40, or a second tower across the river, because the day looks incomplete without a postcard. Incomplete was the point. A short walk and an early coffee is a finished afternoon if that is what you asked for.
A pasteable crowd brief
Paste this before you ask for morning, afternoon, and evening on every day. Fill the brackets with your minutes, your names, and your dates. Keep ticket logic, door times, weather, closure swaps, must-do order, slack, rest days, and kid windows on the briefs linked below. Ask for a day-by-day that shows the famous stop's clock time, whether that clock time is inside a refused window, and what you do if the line is already past the cap when you arrive.
Trip: [cities / region], [N] nights, [dates, and which days are weekend] Travelers: [who], crowd tolerance: [impatient / will wait once / mixed] CROWD HARD RULES: - Wait cap, default: [15] minutes standing in a line or a packed entry. Then abort. - Wait cap, one exception: [named must-see] up to [40] minutes, only at [opening wave / after 16:30], never [12:00-14:30] - Skip-if-packed (famous, optional under pressure): [tower], [miradouro], [old town main square] - Must-see even if busy, but only at a shoulder: [gallery or landmark], first entry wave or late shoulder, not midday - Refused peak windows: no famous stops [12:00-14:30]; no Saturday icon before [16:00] unless it is the opening wave and the wait cap still holds - Weekday vs weekend: [Tuesday market OK at 10:00; same street Saturday is skip]. If a date is Sat or Sun, move icons off that day or cut them. - Abort replacement: a named neighborhood walk within [15] minutes on foot of the aborted stop. Not a second landmark. Not "free time" the model may fill. - Density I will not enter: if the plaza is already shoulder-to-shoulder at arrival, leave even when the formal line looks short. - Photos: one shoulder attempt is enough. Do not schedule a second visit the same day to "get the shot." MUST NEVER ON CROWD DAYS: - Do not put a famous stop in the refused peak window because we are nearby - Do not treat skip-if-packed names as musts - Do not copy a weekday hour onto Saturday - Do not replace an abort with another queue across town - Do not call a crowd abort a rest day and then pack every other square - Do not delete wait caps or skip lists in later revisions to fit more sights - Do not use "avoid crowds" as a substitute for the numbers above LAYER WITH OTHER BRIEFS (paste short lines, do not reinvent them here): - Timed entries (the ticket slot, not the wait cap): [paste] - Opening hours (when the door opens, not when we agree to be there): [paste] - Backup plans (closures, strikes, rain swaps, not a long line): [paste] - Weather and season: [paste] - Must-dos (what we care about, separate from skip-if-packed): [paste] - Open time (slack after a stop): [paste] - Rest days (empty squares, not a crowd-light morning): [paste] - Kids (ages, naps, whether a child will stand in any line): [paste] OUTPUT I WANT: - Day-by-day with clock time on every famous stop - Weekday or weekend labeled on those days - Wait cap visible next to the stop - Abort walk named on any day with an icon - A line that says whether the clock time sits inside a refused window - Checklist: no noon lineup; skip-if-packed not treated as mandatory; Saturday not a copy of Tuesday
One follow-up catches the failure early. Ask the model to list every day that contains a famous stop, the clock time, weekday or weekend, the wait cap, and the abort walk. If a row shows a cathedral at 12:30, or a Saturday copy of a Tuesday hour, or "free time" where the abort walk should be, the brief did not stick. Send that day back by name. Ask for a crowd repair. Do not ask for a new trip.
For the lines this page refuses to swallow, use the briefs that already exist: timed entries, opening hours, backup plans, weather and season, must-dos, open time, rest days, and kids. The generic itinerary guide is the wider stack if the whole draft is still a postcard.
Where this stops and the other guides start
A timed entry is the ticket in your pocket: a 9:10 slot, a barcode, a rule about being late. It does not decide whether you should have accepted a 13:00 slot because it was the one left on the calendar. If the only remaining ticket sits inside a window you already refused, the crowd brief says skip, even when the ticket is real. Keep slot math on the timed entries guide. Keep the refusal here, in minutes and clock times.
Opening hours tell you the door is open. They do not tell you the plaza is kind. A site can open at 9:00, close at 19:00, and still be a lineup at 12:20 with a perfectly legal door. Use the opening hours guide so you do not arrive to a locked gate. Use this page so "open" is not treated as "go now."
Backup plans are for the day the thing is not available. A strike, a sudden closure, rain that makes the outdoor plan pointless, a missed train that kills the slot. A crowd abort is the opposite situation: the door is open, the ticket might even be valid, and the line or the density is past what you will do. Do not let the model reuse your indoor rain swap as the answer to a long line, especially if that swap is another queued museum on the far side of town. Closures live on the backup plans guide.
Weather and season change how many people show up, and they change whether midday is miserable for reasons that are not only the queue. A hot June plaza and a gray November plaza are different trips. The weather and season guide holds that. This page still applies on a perfect sunny day. Sun is when the noon lineup is worst, not when the rule expires.
Must-dos answer a different question: what you care about enough to protect. A crowd rule can demote a famous stop without erasing that ranking. You might care deeply about one gallery and still refuse a tower that every sample day includes. If the model collapses both into "top sights," it will spend your one patient hour on the tower. Rank on the must-dos guide. Mark skip-if-packed here.
Open time is slack after a stop so a late exit does not wreck the next thing. Filling that gap with another icon misses the point, and leaving a line is a different decision. Pair a buffer from the open time guide with the abort walk here. The buffer keeps the afternoon from shattering. The abort walk is what you do instead of the queue.
Rest days are empty on purpose: laundry, a sleep-in, a day with no tickets. A morning you kept light because the landmark was packed is not that day. If the model counts the abort as your rest day, every other square gets packed to "make up for it," which is how you end up with two noon lineups later in the week. Keep rest pins on the rest days guide.
Kids add a harder cap that this page should not reinvent. A four-year-old who will not stand for ten minutes changes the wait number, and nap windows change which shoulder is even available. Put ages and naps on the kids guide. If you are traveling with children, paste both. A crowd brief alone will still invent a noon monument and call it family-friendly.
A weekday that refuses the noon lineup
ILLUSTRATIVE sketch only, not a real booking and not a city guide. Five nights, one base, two adults who will wait once and otherwise leave. Day 1 is arrival, no icon. A short loop on the streets around the lodging, done before dinner, capped around an hour. The model is not allowed to "upgrade" that loop into the old town square because the square is famous.
Day 2 is a Tuesday. The one must-see gallery is the opening wave, out by about 10:30, then the named neighborhood on the same side of the river. If the line at the door is already past 25 minutes, skip the gallery and take the neighborhood. Do not slide the gallery to 13:00 to compensate. Compensation is how the noon lineup returns wearing a different sentence. Day 3 holds the viewpoint for after 16:30, or drops it. Midday is the market as a walk-through, not as a queue for a terrace. If that date were a Saturday, this viewpoint moves off the day entirely. The Saturday plan is a neighborhood and a long lunch, not a delayed postcard.
Day 4 is the one patient hour. A single building you will wait up to 40 minutes for, at the first wave, with the abort walk written underneath: the residential grid behind it, not the other famous church. If you abort, the day is finished. No 17:40 retry. Day 5 is the rest day from the other brief, actually empty, not a crowd-light morning with three small sights. When a model adds the cathedral steps at 12:30 "because you will be nearby," this sketch rejects the line. It looks a bit dull in a list. It is the difference between seeing the neighborhood at 9:10 and seeing the back of someone's umbrella at 12:40.
The common default, for the same five nights, is gallery at 11:00, tower at 13:30, viewpoint at 17:00 "for the light," and the neighborhood if energy remains. Energy goes to the lines. The difference is a crowd brief that still has its numbers on day four of the chat, not only in the first message.
Ask for clock times and wait caps on the sheet before anyone picks restaurants. If the famous stops still sit inside 12:00-14:30, clear those hours by name. Cafes can wait. A perfect lunch reservation next to a noon lineup does not fix the lineup.
Red flags in the draft
A famous stop between 12:00 and 14:30, with the early visit mentioned only as an option. The option is not the plan. Look at the clock time in the day body.
"Avoid crowds" still sitting in the notes, and no minute cap anywhere. The model agreed with your mood and scheduled the postcard anyway.
Every starred name treated as a must. The tower, the square, the viewpoint, the second church. Your skip-if-packed list is missing, or it appears once at the top and then every day includes those names.
Saturday written as Tuesday. Same hours, same icons, a word like "lively" added to the market. Lively is doing a lot of work there.
The neighborhood scheduled after the monument. You arrive with the spillover. The brief wanted the lanes first, or instead.
The abort is another queue. You leave the dome line and the replacement is a palace across town with its own entry wait. That is a second landmark, even if the sentence says "quieter alternative."
A crowd abort renamed as rest. The model counts the light afternoon as your empty day, then stacks icons on the other days to recover the "lost" sightseeing.
A ticket inside a refused window. The 13:00 slot was available, so it became the plan. Availability is not tolerance. Send it back even if the barcode is real.
Rules that lived only in the first message. By revision three the wait cap is gone, the skip list is gone, and the viewpoint has returned at sunset because the thread got long. Paste the crowd brief again until the clock times and the caps are still on the sheet.
Change the hour without starting over
A noon lineup rarely needs a full regenerate. Regenerating tends to invent a new icon for the same hour and drop the names you were willing to skip. Stay in the thread and edit the broken day.
Name the square first. "Day 2 puts the tower at 12:40. That is inside the refused window. Wait cap is 15 minutes and this tower is skip-if-packed. Cut it or move it to the opening wave if the door allows, otherwise delete it. Abort walk stays the lanes behind the market. Do not add a second monument. Do not move the rest day." Specific beats "make it less crowded," which the model will answer with the word "stroll."
Restate the cap, the skip list, and the weekend rule in that same message. Ask to see them on the revised day, next to the clock time. If the numbers vanish, the edit did not stick. If a new ticket slot lands at 13:00, reject the slot even when the rest of the day looks cleaner. Keep closure backups and kid windows untouched. Those are other briefs, and a crowd fix should not eat them.
The refine guide has the keep, change, and drop pattern. Add one crowd line to it: "Keep wait caps and skip-if-packed names on Days [2-4]; no famous stop in [12:00-14:30]; Saturday is not a copy of Tuesday; abort walks stay neighborhood, not a second queue." When one afternoon ignored the line, repair that afternoon and check the next day so the model did not "make up" the stop tomorrow at noon. Do not regenerate the week.
How this sits with the other briefs
This page is the crowd pattern: minutes you will wait, hours you refuse, famous stops you will skip when they are packed, and a nearby walk for the abort. The other pages keep tickets, doors, weather, rankings, slack, rest, and children from being flattened into "avoid crowds."
Timed entries hold the barcode. Opening hours hold the door. A legal ticket at a legal open hour can still be a noon lineup, and that is the case this brief exists to catch.
Backup plans cover closures, strikes, and rain swaps. A long line with the door open is not a closure. Weather and season change the kind of day. They do not retire the wait cap when the sky is clear.
Must-dos say what you came for. Skip-if-packed says which famous names lose when the plaza is full. Open time is slack, not a secret slot for a second icon. Rest days stay empty even if you aborted a line yesterday. Kids sets whether anyone in the group can stand in a line at all.
Generic AI itinerary work is the fuller stack before the first draft, when the problem is bigger than crowds. Refine is the edit after one day ignored the cap.
When a preference survey is worth it
If you are willing to keep pasting the crowd brief, a general assistant plus the minutes and the skip list above is enough for a lot of trips. The sheet is the product. Clock times you can check with your own eyes matter more than a nicer paragraph.
The structured path starts to matter when the chat keeps parking the famous stop at noon after the cap was explicit. Or when skip-if-packed names and the Saturday rule vanish by revision four. Or when you want a researched day-by-day that still holds those constraints without teaching them again in every new thread.
One option with that shape is Dople (dopletech.com). A short preference survey can carry crowd tolerance with pace, food, and lodging. You compare trip options shaped by that profile, and you can buy a Trip Project: a researched day-by-day meant to hold the preferences across revisions, so you are not stuck with a noon lineup the model keeps restoring. The account and the survey start are free. Full itineraries are paid. No free itinerary is tucked behind checkout. More on the shape of that work: a comparison with ChatGPT and what the product is.
Public samples show ILLUSTRATIVE day-by-day shapes only. They are not live bookings, and they are not free Trip Projects.
If that survey path sounds useful, sign up once, when you are ready. If it does not, use the brief above in whatever chat you already have open. The wait cap works either way.
What does planning an AI itinerary around crowds mean?
It means giving the model a wait cap in minutes, the peak windows you refuse, which famous stops you will still visit at a shoulder hour, and which you will skip if they are packed, plus a nearby walk for when the line is already too long. Without that, drafts park the landmark at noon and call the neighborhood optional.
Why do AI drafts drop you at the famous stop at noon?
Sample days put the monument in the middle. "Avoid crowds" does not beat that shape. A minute cap, a skip-if-packed list, and a note that Saturday is not Tuesday give the model fewer ways to schedule the lineup and then apologize under it.
What belongs in a crowd brief?
The wait cap, refused hours, must-see versus skip-if-packed names, shoulder versus midday, weekday versus weekend, a named abort walk, and short pointers to timed entries, opening hours, backup plans, weather and season, must-dos, open time, rest days, and kids.
Should I regenerate when the AI ignores a skip-if-packed rule?
No. A new draft often finds another famous stop for the same hour and loses the list. Name the day, restate the cap, move or cut the stop, keep the abort walk. The refine guide is the pattern for editing in place.
Are the public samples real bookings?
No. Samples are ILLUSTRATIVE examples of itinerary shape. Real Trip Projects are researched per traveler after the preference survey. More context sits on the personalized travel itineraries FAQ.