All posts

Draft — not published

Unlisted on the blog index and marked noindex, so it won't show up in search. Set status: published in the file's frontmatter when it's ready.

Guides·3 min read

Five ways to spot an AI itinerary that will fall apart on the trip

AI trip planners have a specific failure mode: the output looks more polished than a human's, and is wrong in ways you can't see from the page.

Here's how to check one in about five minutes. This applies to any of them, ours included — and honestly, run it on ours. We'd rather you caught something than trusted it blindly.

1. Google two restaurants at random

Not the famous one. Two ordinary ones, mid-week, from different days.

You're checking three things: does it exist, is it in the city the plan says, and is it permanently closed. This single test catches the most common and most embarrassing failure, which is a real restaurant with the right name in the wrong city. Portland has great restaurants in both Maine and Oregon.

If two random picks check out, the plan was probably verified. If one is a ghost, assume the rest is decorative.

2. Check the travel day

Find the day you change cities. Look at breakfast.

If breakfast is in the city you're arriving in, the plan doesn't understand that you wake up somewhere. It's a small tell with big consequences — it usually means the whole day is assembled from the destination's list with no sense of the journey, and the timings on it are fiction.

3. Add up one afternoon

Take any day and read the times literally. 2:00–4:00 at a museum, 4:15 at something across town — is that fifteen-minute gap actually possible? Do it for one day only; you'll know immediately whether transit time exists in the model or not.

The tell for a bad plan: every activity is exactly one or two hours, and every gap is exactly fifteen minutes. Real days aren't that tidy.

4. Look for the day that's too good

Six things, all excellent, all clustered, ending with a 9pm dinner. That's not a day, it's a highlight reel. A real plan has a day that's deliberately lighter, and a travel day that's mostly travel.

If nothing in the week is quiet, the planner is optimizing to impress you rather than to be survivable on day five.

5. Check what it says about itself

Does the plan tell you it's AI-generated? Does it say to verify hours and bookings?

A tool that hides its own uncertainty is making a choice about who absorbs the risk of being wrong, and it isn't choosing itself. We put that disclosure on every page that shows a plan — not because it's legally required boilerplate, but because the honest version of this product tells you what it does and doesn't know.

What good looks like

A plan you can trust is usually a bit less impressive on the page: fewer perfect days, more explicit gaps, occasional notes about closures, and a travel day that's honest about being mostly travel.

We build tripcoord that way on purpose. Every venue goes through a check against real place data before it reaches you — right city, real business, not closed — and if something fails we regenerate that piece rather than shipping it. It makes our builds slower and costs us more per trip.

It also means that when you're standing on a street at 8pm, the restaurant is there.

[CTA → try a free build]

Planning something with other people?

tripcoord builds the day-by-day plan, checks every venue is real, and keeps the group in one place. First itinerary is free.

Build a trip free