Skip to content

Planning a trip with ChatGPT: precise prompts that avoid generic recommendations

Asking an AI to "plan a trip to São Paulo" rarely produces a memorable itinerary. According to experts interviewed by CNA Luxury / The New York Times, the rule is simple: a generic question gets a generic answer. To get useful recommendations, you need to feed the bot personal context and then verify every suggestion. Here is how to structure your prompts, with rewriting examples and verification criteria.

Why vague prompts fail

A prompt like "what should I do in New York?" lets the AI draw from the most common lists: Times Square, Central Park, a famous museum. The result is not wrong, but it fits no one. The experts cited by the source recommend providing a detailed travel profile: tastes, goals, preferences, desired pace, rough budget, dietary or mobility constraints. The more precise the brief, the more precise the answer can be.

Rewriting examples

Before / after

Vague prompt: "Where should I eat in São Paulo?"Precise prompt: "I'm spending four days in São Paulo in October. I like regional Brazilian food, I prefer neighborhood spots over fine dining, I eat early and I'm vegetarian. Give me ten restaurants grouped by neighborhood, with cuisine type and a price range."

Vague prompt: "Make me an itinerary for New York."Precise prompt: "I'm in New York for five days, I walk a lot, I like jazz, architecture and food markets, and I want to avoid long lines. Build a half-day-by-half-day plan, grouping places by proximity."

Ask the bot to ask questions

One tip reported by the source: ask the AI to pose five follow-up questions to clarify the request before producing the itinerary. This forces the model to identify missing information — budget, dates, pace, interests — instead of assuming on your behalf.

Split rather than stack

For São Paulo, the author asked for four separate lists — activities, restaurants, nightlife, accommodation — rather than a day-by-day program. This split makes verification easier and lets you compare options before building the sequence yourself.

Feed the bot other sources

The source recommends integrating friends' advice or other research into the bot. AI does not replace a human recommendation: it organizes and complements.

Limits on information access

Bots do not have access to the entire internet. Many sites block their crawling or have specific agreements. According to the source, ChatGPT and Gemini have full access to Reddit, unlike Claude. The author also reports that Claude and Gemini could not search Eater directly, while ChatGPT could. These observations date from the article's period and may change with commercial agreements.

Reported errors: what they mean for you

The source reports several concrete mistakes. Gemini recommended a restaurant and a music venue that had already closed, an error the bot acknowledged. Gemini first suggested a hotel, then retracted after checking Google reviews that flagged construction noise. Claude hallucinated by recommending two Brazilian bakeries that were actually a parking lot and a Japanese restaurant. The author notes that ChatGPT made no errors he spotted, without guaranteeing there were none. These tests were run by the author, not by an independent evaluation: results may vary by account and version.

Verification criteria before booking

  • Existence and opening: check the address and hours on an independent source.
  • Recent reviews: consult Google reviews or an equivalent platform to spot closures or construction.
  • Direct access: if the AI cannot consult a specialist site, open the site yourself.
  • Geographic coherence: check that grouped places are actually close.
  • Price and booking: confirm on the official site, not only in the AI's answer.

Organizing the conversation

Creating a dedicated project on ChatGPT or Claude lets you keep a travel conversation and refine it later. The source notes that Gemini has no exact equivalent to the Projects feature. A dedicated project avoids mixing contexts and makes follow-ups easier.

Key takeaways

  • A generic prompt produces a generic answer: give a detailed profile.
  • Ask the AI to pose follow-up questions before generating the itinerary.
  • Prefer separate lists by category over a day-by-day program.
  • Verify each place independently: existence, hours, recent reviews.
  • Paid versions went deeper and faster according to the author, but verification remains essential.

This article is published by Roboto, a platform for generating texts, images, videos and voiceovers with AI. Roboto performed no tests on the tools mentioned; the observations reported come from the cited source.

Source

Planning a trip with AI? Here’s how to do it better

Related reading