9:14am, I'm standing in a parking lot the chatbot told me to park in, looking at a sign that says the lot doesn't open until 10, and already the day is off-script in a way the day itself hadn't even started yet.
The rule was simple and I made myself stick to it: whatever the AI itinerary said, I did, exactly as written, no substitutions, no checking Google reviews first to see if it was a trap. Budget was $50 total, food, activities, parking, everything. One day, my actual city, no cheating by picking a town I already know is full of good options.
I typed one prompt the night before: plan me a full Saturday, $50 total budget, starting 9am, must include something outdoors, something food-related, and one thing I've never done in this city despite living here six years. Then I closed the laptop and didn't look again until morning.
The Question
Can a chatbot actually plan a decent, specific, budget-real day out in a mid-size city it has no live access to, or does it just generate plausible-sounding nonsense that falls apart the second you try to follow it literally?
Everyone tests AI trip planning by then quietly fixing all its mistakes before they go. I wanted the version where I don't get to fix anything, because that's the actual test of whether the plan works.
The Setup
Budget: $50 flat, tracked to the dollar, including parking and any tips. Rule one: no substitutions, if a place is closed or wrong I have to find the closest possible literal interpretation of the instruction, not just pick something better nearby. Rule two: no pre-checking reviews or hours before I leave the house. Rule three: I'm allowed to ask the AI follow-up questions if something goes wrong, but I have to follow whatever it says next too.
The itinerary it gave me: 9am parking at a specific downtown lot, a farmers market that runs Saturdays, a specific coffee stand inside it, a 45-minute walk along the river trail, lunch at a specific taco truck, an art supply store it called "a great low-cost activity spot" (unclear what that meant), and a sunset spot at a park on the east side.
What Happened
The parking lot didn't open until 10, which meant I circled for a metered spot two blocks over and burned $4 I hadn't planned for before the day had technically started. Not the AI's fault exactly, it had no way to know the lot's seasonal hours, but it's the first sign this was going to require some real-time interpretation.
The farmers market was real, running, and genuinely good, and the specific coffee stand it named exists and the coffee was excellent, $5 for a pour-over that was better than what I usually get near my house. First real point in the AI's favor, and it was oddly satisfying to have it be right about something that specific.
The river trail walk was fine, pleasant, free, exactly as described. The taco truck it named for lunch, though, had closed and been replaced by a completely different truck at the same spot, selling Korean corn dogs instead. Rule one meant I had to eat there anyway since it was the literal location named. Two corn dogs, $11, genuinely delicious, better than the tacos probably would have been, and a place I never would have found on my own.
The art supply store as "a great low-cost activity spot" turned out to mean something specific once I got there: a paint-your-own-pottery corner in the back I didn't know existed, $18 for a small mug I could glaze and pick up later. This is the moment that actually surprised me, because the AI's vague phrasing turned out to be describing something real and specific that I, a six-year resident, had never noticed on a street I walk down constantly.
By the sunset park, I was at $44 with one stop left and no clear budget line for how to spend the last $6. I bought a lemonade from a cart at the park and sat there with $1 left over, watching the sunset it had, somewhat presumptuously, described in the original prompt response as "a good place to reflect on your day," which felt like the AI narrating its own itinerary a little too confidently.
The AI is a great idea generator and a mediocre logistics planner, and the whole fun was not being allowed to tell the difference until I got there.
What Went Wrong
The closed taco truck is the real weak spot: an AI planning a specific day in a specific city has no live knowledge of what's currently open, and following its literal instruction when it's wrong could easily have led somewhere worse than corn dogs. I got lucky.
The parking lot hours mistake cost real money, small but real, and it's the kind of error that compounds if you're planning a longer or more expensive day where a wrong assumption about hours or availability could blow the whole budget, not just $4 of it.
The itinerary had zero slack built in for anything going wrong, which meant every deviation, closed truck, unopened lot, ate directly into either the budget or the schedule with no cushion.
The Verdict
Would do this again, specifically as a low-stakes way to see your own city differently, not as a reliable logistics tool. The pottery stop alone justified the whole exercise, that's a genuinely good recommendation I'd never have found browsing lists of things to do.
Good for anyone bored of their own weekend routine in a city they think they already know well. Bad for anyone with a tight schedule, a hard budget with zero flexibility, or plans that depend on a specific place actually being open when they arrive.
The honest verdict is that the AI is a genuinely good idea generator and a mediocre logistics planner, and the fun of this experiment was entirely in not being allowed to tell the difference in advance.
Steal This
- Ask for one thing you've "never done despite living here X years" specifically. That constraint is what produced the actual best moment of the day.
- Build in a $5-10 buffer beyond whatever budget you give it. Hours and closures are the thing it can't know, and something will be wrong.
- Don't pre-check the plan if the whole point is discovering something new. The moment you start correcting it in advance, you're just planning your own day with extra steps.
- If a specific spot is closed, don't skip it, go to whatever replaced it. That's where the actual surprises happen.