Using the LLM to plan a trip
Creating a schedule, compiling review website data, verifying local customs and more
Is it better than Yelp & Trip Advisor?
I used ChatGPT to map out five days in Manhattan. The original goal was to see a Broadway play. I created a project called NYC Trip and saved important trip details (dates, flight number, hotel, confirmation numbers) in the instructions. In prompting, I could just say "how far from the hotel?" or ask it to remind me what time the flight lands.
It assisted several ways. It kept a schedule I built, tracked which attractions were slotted/rejected/left to discuss. I ran deep-research prompts to get live data as the date got closer.
So little time, so many fake subway directions!
There are a few areas where I mistakenly believed the LLM had more knowledge than it had. It doesn't know it doesn't know, so it can't really "tell" you that it's faking it. I wrote about drift, hallucinations and the mirror, here.
- Even though I updated a hop on/off bus company in the project instructions, for the rest of the project, it kept recalling the old bus company intermittently. I never found a workaround for this, so I had to remember.
- It's good at compiling data for restaurants, pulling data from Yelp, Trip Advisor, Google reviews, but not good at grabbing hours, menus and if the restaurant is still open.
- It was terrible at navigating the NYC subway. We found the subway challenging on maps, so we tried the LLM. It confidently gave us the wrong directions.
- I got analysis paralysis. I got "sucked-in" to reviewing one more restaurant, one more lounge; there's always more you "can" do.
The most valuable component was how it kept track of the schedule, by sanity checking activity time(you're too busy here, you have time available there.) Note that I was only able to save this because I have memories turned on. Once the trip was over I went in and deleted all NYC related memories.
Below are some of the types of prompts I used in this process. For more information on prompt building, start here.
Keeping a schedule
In the project instructions, I had a calendar structure. I told it to keep the days as column headers and the time of day as row headers (breakfast, day, lunch, afternoon, dinner, evening.) As I made decisions, I'd have the LLM update the schedule. "Add Joe's pizza to Wed lunch and save." At any time, I could recall the calendar.

Preview: Schedule snapshot managed by the LLM (zoomed-out view)
Through prompting, it helped me keep a realistic schedule. "There's no way you are going to do a sightseeing cruise, go out to lunch and do the museum all before Broadway. It will be too hot and you will need a chance to rest."
Analyze schedule as defined in project instructions. Focus is putting an estimated start time/end on each activity to see where there are large blocks of time (over 2 hours,) or where we are too crunched (either overlapped or no time to rest or account for delays. Output analysis by day/time block.
This project maintains a structured daily trip calendar covering the dates June 23–28, 2025.
The calendar uses a matrix format:
- Columns = Day/Date (6/23–6/28)
- Rows = Time chunks:
- Breakfast
- Morning
- Lunch
- Afternoon
- Dinner
- Eve
Each cell may include activity suggestions, confirmed bookings, restaurant targets, fatigue/weather notes, or “left open.”
Usage rules:
- All proposed activities must be slotted into time/date blocks and marked as:
- Planned
- Booked
- Confirmed
- Canceled
- Left open
- System should maintain continuity of the calendar and return updated versions on request.
- Do not assume activities span multiple blocks unless explicitly stated.
- Never overwrite existing confirmed entries unless instructed.
All planning, attraction reviews, and restaurant prompts must ultimately feed into this schedule.
Compiling Restaurant Reviews
I wasn't planning to use it for finding restaurants, but I wasn't having any luck on Yelp. A query of the LLM led to my learning that the locals in Manhattan aren't big Yelp fans, which led me to finding a visitingNYC Reddit. I started using the LLM to compile data from Yelp, Trip Advisor, Google Reviews and Yelp sites. I had it bring back star ratings, menu highlights that kept getting repeated or sentiments of reviews.
Find me 2-3 locally loved, classic breakfast diners within 0.5 miles of Radio City Apartments (142 W 49th St). Prioritize places with consistent local reviews over tourist hits.
Pull data from Yelp, TripAdvisor, Google, and any NYC-specific forums.
Return:
Name and address
Summary of top 3 highlights from reviews (weigh by frequency, not stars)
Summary of top 3 complaints (same rule)
Estimated cost for breakfast for two
Menu sample or specialties called out by reviewers
Which experiences are worth it?
Figuring out what to do with our limited time was challenging. There is so much to do! We had to be picky and keep in mind the time of year, time constraints, physical limitations and budget. The LLM helped me learn as we went along. I kept adding things to the "to do" list, and it kept estimating how long the activity would take, including transportation time. At one point, I had eight full days of activities without a break. (We had almost 5 days and would need breaks.)
I had the LLM run through a variety of exercises to compile data from Trip Advisor, Google Reviews and r/visitingNYC. I wanted it to grab a consensus from users on how each experience was viewed. It could also grab hours, costs and whether VIP experiences were worth it. It brought back real information on bringing cash or if restrooms were clean. It read hundreds of user reports from 3-4 sites and outputs, "Users report being able to sit for 1/2 hour before the venue opened but there was no shade. Many recommend bringing a bottle of water and wearing a hat."
Bring me a side-by-side comparison of the top 3 observation deck experiences in NYC. Include:
Name, location, and neighborhood
Height of the deck (in feet and stories)
Admission cost for two adults (standard tickets, not VIP)
Vibe or style of the experience (e.g., flashy, quiet, historical)
Unique features (glass floors, open-air sections, sunset views)
Crowd level and best time to go
What locals say — top 3 pros and top 3 cons, weighed by frequency
Weather dependency or indoor/outdoor access
Accessibility (e.g., mobility, stroller, sensory friendliness)
Any combo ticket or bundle options worth considering
Use sources like official sites, Reddit threads, NYC blogs, Google, Yelp, and TripAdvisor. Focus on *actual visitor commentary*, not promotional fluff. I want summary judgment, not raw links.
Deep Research one-week pre-trip
Most of the planning had been done. We decided to keep the schedule loose but had a ton of information saved from all the research. A week before we left, I ran a deep research prompt to verify NYC restaurants by meal type and itinerary location; assess worthiness, avoid filler, and check for missed experiences. (This prompt uses the schedule I created to see what we might be missing.)
The output of the prompt was a 19-page Word document that I digested and whittled down to something manageable we could use in NYC. It confirmed hours, restaurant was still open and provided recent comments.
Goal:
Identify 5 restaurants each for breakfast, lunch, and dinner near or accessible from Radio City Apartments (Midtown West) that deliver excellent food experiences under a realistic price ceiling. Include 1 additional review of Trattoria Trecolori to evaluate whether it remains a top pre-theater dinner option.
Meal Price Thresholds (for 2 people, pre-tip):
| Meal | Low-End Target | Mid-Tier Target | Avoid If Over |
|----------|----------------|-----------------|----------------|
| Breakfast| $30–$35 | $40–$45 | $50+ |
| Lunch | $35–$45 | $50–$60 | $65+ |
| Dinner | $60–$75 | $85–$100 | $125+ |
Cuisines to Prioritize:
- Italian (fine dining and red sauce)
- Jewish deli classics
- New York pizza by the slice
- NYC dirty water hot dogs
- Bacon Egg & Cheese (BEC) bodega-style sandwiches
- Iconic NYC desserts (e.g., black & white cookie, cheesecake)
- Burgers (top-tier NYC-style only)
- Soup dumplings (Chinatown only)
- Halal cart favorites (only if clean and legit)
Explicit Exclusions:
- Vegetarian/vegan-focused spots
- Gluten-free or allergy-oriented restaurants
Include:
- Any “Soup Nazi”-style spots worth the hassle
- Food carts with real followings (locals + hygiene pass)
- Iconic experiences even if simple
Fields to Include for Each Entry:
- Name
- Location (full address + walk/subway info)
- Sort Category: Quick Bite / Sit-Down / Leisurely Meal
- What Locals Say (esp. NYers, not tourists)
- What to Order (house specialties)
- Meal Cost for 2 (estimate, no alcohol)
- Reservation Needed?
- Late-Night Viability?
- Why It Made the List
- Tags for NYC Food Experience (pizza slice, burger, etc.)
- Heads-Up: Strict service, wait times, cash-only, etc.
Additional Query: Worthwhile Off-Site Meals
For any location on the itinerary that is outside Midtown (e.g., Coney Island, SoHo, Upper West Side, etc.), identify 2–3 highly regarded lunch or dinner spots *near that area*. Prioritize places that are:
- Local favorites or iconic to the neighborhood
- Worth staying in the area to eat at
- Not tourist traps or forgettable filler
For each restaurant, include a quick note:
> “Worth staying in the area for” OR “Good, but Midtown has better.”
Do *not* include suggestions solely based on proximity—only highlight if it’s legitimately a worthwhile food stop.
Experience & Itinerary Crosscheck:
Evaluate the current NYC itinerary (June 23–28, 2025) and compare it against:
1. Highly rated or iconic NYC experiences commonly recommended by locals or seasoned visitors (not just tourists)
2. Items we've discussed or considered but haven’t scheduled
3. Any limited-time or seasonal events happening during our travel dates that are actually worth it
Return a prioritized list of 5–8 total experiences. For each, include:
- Name
- Concise Description (What It Is / Includes)
- Estimated Time Required
- Price (per adult)
- Location & Distance from Hotel (walk or subway)
- Local & Visitor Impressions (honest, not fluffed)
- Why It’s Worth Doing (if it is)
- If it's skippable, say so and why
The goal is to verify that our planned itinerary holds up, avoid missing anything exceptional, and cut or replace lower-value items if something clearly better exists. Skip all generic tourist filler.
VERIFICATION REQUIRED:
Do not include any restaurants, food carts, or experiences unless they can be verified through **recent reviews (2024 or later)** from **reputable sources or aggregated review platforms** (e.g., Google Maps, Yelp, Eater, Time Out, The Infatuation, etc.).
Do not invent or assume a restaurant exists based on cuisine type or neighborhood. If data cannot confirm a place exists, is still open, and maintains current relevance, **exclude it completely**. Flag anything that recently closed or appears borderline.
If two sources conflict, default to more recent or local-reviewer consensus.
Real time geo locating
While in NYC, we kept getting lost. I could take a picture, feed it to chat, and tell it, "Get me to hotel from here." I'd get back - turn left, walk 1 block to subway, etc.
"I'm at Grand Central Station. Help me find the bar." I'd get back: First, it's Grand Central Terminal, now. Second, go left, walk up the ramp and turn right.
"Maps tells me I'm standing in front of Mood Fabrics but I don't see it. Help!" I got back: Turn around!