"AI trip planner" is a phrase that's earned some side-eye. Too many of them generate confident, generic itineraries, plausible-sounding days that fall apart the moment you check whether two stops are actually near each other, or whether the museum is open on a Monday.
InChina's planner is built differently, in two specific ways: it works in two stages instead of one, and it draws from real local guides instead of inventing places. Here's what actually happens between "I have five days in Yunnan" and a plan you can walk.
Stage one: the route, before the detail
The first thing the planner does is not write your days. It reads. You give it the basics, a city, how many days, your pace, and what you care about, whether that's food, history, nature, or photography, and the first stage drafts the route: which places, in which order, across which days.
This stage stays deliberately lean. It's deciding the shape of the trip, the skeleton, without yet filling in every meal and tip. Getting the order and the geography right first is what keeps the later detail from collapsing. A good route means you're not crossing the city twice in a day or backtracking to a stop you already passed.
Stage two: filling the days in
Once the route holds together, the second stage does the enrichment, going day by day to add the food spots, the local tips, the photos, and a smart order for getting around. This is where a bare list of place names turns into something you'd actually follow: real opening hours, what's worth eating across the street, what locals do at a given hour.
The reason this two-stage split matters is that it mirrors how a good human planner works, you don't decide what to eat for lunch on day three before you know whether you're even in that neighborhood on day three. Route first, detail second. It's the same pipeline our editorial team uses for the guides we publish by hand.
Drawn from real guides, not invented
The part that separates this from a generic chatbot is the source. The planner drafts from the same library of local guides that human writers built, places picked by people who actually live in those cities. So when it suggests a stop, it's pulling from a real, walked guide, not generating a place that sounds right and may not exist.
That's also why it isn't trying to plan all of China at once. It plans well where there are real guides to plan from, and the library grows as writers add more.
It drafts. Then it steps back.
Here's the honest boundary, and it's the most important part. The AI plans the route, it doesn't decide what's worth seeing. It hands you a draft, and then it gets out of the way:
- Drag to edit. Swap a place you don't want, reorder a day, drop a stop entirely. The AI doesn't push back or "correct" you. It's a first draft, not a verdict.
- Saved offline. Once a plan is drafted, it lives on your device. Itinerary, maps, notes, all work in flight mode, in a mountain village, on a sleeper train with no signal.
- Yours, privately. Your itineraries and searches are yours. No ads, no "recommended for you" pressure steering your trip toward someone's booking commission.
The goal was never an AI that tells you where to go. It was an AI that handles the boring scaffolding, the ordering, the timing, the "is this even near that", so the actual decisions stay yours.
Try it, then make it yours
The best way to understand it is to run it: pick a city, a few days, a vibe, and watch the two stages draft a sample plan from real itineraries. Then do the thing the demo can't, edit it, save it, and carry it offline into a place where the signal drops and the plan still works.
Plan your trip in the app.