Photo via Toronto Star

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

Read the original at Toronto Star ↗
Cite

Seth Kugel The New York Times. (2026, September 24). Planning a trip with AI? Here’s how to do it better. Toronto Star. https://www.thestar.com/life/travel/planning-a-trip-with-ai-here-s-how-to-do-it-better/article_42de618f-e298-4d47-a3fc-08825fe679ed.html

Dated when The Aporia collected the article, usually within hours of publication. Check the original for its exact date and byline.

Toronto Star · collected 2026-09-24 · by Seth Kugel The New York Times

Quick Summary

The article explores how artificial intelligence can improve travel planning by providing tailored advice for specific cities like São Paulo and New York. It highlights that AI requires detailed prompts to avoid generic responses and to prevent companies from gaming the system with "generative engine optimization." The piece suggests compiling a personal travel profile including tastes, goals, and preferences to get more customized results from chatbots.
Written by AI on 2026-09-24, from this article's own text only (the story summary below draws on all the coverage).

AI analysis

Story summary

Seth Kugel from The New York Times explores how artificial intelligence can enhance travel planning despite its current limitations. According to Ethan Mollick, a professor at the Wharton School of the University of Pennsylvania, AI chatbots are not infallible oracles but require better prompting to deliver optimal results. Kugel tested this by creating fictional five-day trips for São Paulo and New York, using three different AI bots: ChatGPT, Gemini, and Claude. He found that paid versions of these tools generally offered more comprehensive and quicker responses compared to their free counterparts. The goal was to demonstrate how even non-experts can use AI effectively if they craft specific and thoughtful prompts, highlighting the potential for AI in personalizing travel itineraries based on individual preferences and needs.

Written for “AI Travel Planning Tips” on 2026-10-31, grounded in this article and the 0 other(s) covering the same event.

Story

📰 AI Travel Planning Tips
Technology · 1 article(s) covering the same event.

Compared with similar articles

Nothing to compare against. No article is close enough to this one for the pipeline to have linked or judged the pair.
How this article was measured signals, leaning charts, publisher record, byline, topics and claims

Signals How these are calculated →

Claims extracted
67
sentences that state something as happening or true
Uncertain
10%
7 of 67 hedged ("reportedly", "may", "according to")
Leaning
not political
takes no side on a contested political question
Publisher signals
64/100
Toronto Star's corrections and hedging, across everything collected from it. Not a measure of accuracy.
Outlets on this story
1
Technology
Same framing elsewhere
1
articles that frame the event the way this one does
Analyzed 2026-09-24 · how these are computed

How this is being covered How these are calculated →

Article leaning vs. publisher reliability
Source leaning vs. consistency

Publisher

Toronto Star · 10694 article(s) · 27 correction(s) detected
Running correction rate · 27 correction(s)
2026-10-26
Correction
2026-10-24
Correction
2026-10-07
Correction
2026-09-24
Correction

Who wrote this

Seth Kugel The New York Times
1 article(s) here · 1 carrying a prediction
🔮 Chatbots may still be a long way from replacing old-school travel planning techniques, but they can create far better results than most of us realize.
2026-09-24 · assertive framing · Planning a trip with AI? Here’s how to do it better
The only article under this byline in the corpus.

Topics

Brazil Rome São Paulo the University of Pennsylvania the Wharton School

Subjects

São Paulo GPE · 4× Chicago GPE · 2× Mollick PERSON · 2× New York GPE · 2× Bon Jovi PERSON · 1× Brazil GPE · 1× Ethan Mollick PERSON · 1× Rome GPE · 1× the University of Pennsylvania ORG · 1× the Wharton School ORG · 1×

Narrative

Mine included my love of “tours with very quirky local guides,” my aversion to international hotel chains and my tendency to choose restaurants based in part on online user reviews written in the language of either the destination (Paris restaurants reviewed in French) or the cuisine (Chinatown restaurants reviewed in Chinese).
framing: assertive · carried by 1 article(s) · first seen 2026-09-24
🔮 Chatbots may still be a long way from replacing old-school travel planning techniques, but they can create far better results than most of us realize.
2026-09-24 · Toronto Star
Planning a trip with AI? Here’s how to do it better · assertive framing

Claims (67 extracted, 7 hedged)

If artificial intelligence can quickly find needles in legal haystacks, crack long-unsolved math problems and create retro 1980s images convincing enough to be Bon Jovi album covers, why is it so bad at planning a week in Rome? asserted
it → find → Rome
Chatbots may still be a long way from replacing old-school travel planning techniques, but they can create far better results than most of us realize. uncertain
most → replace → us
The problem: We’re not helping them enough. asserted
We → help → them
“They’re not oracles,” said Ethan Mollick, a professor who studies AI and its impact at the Wharton School of the University of Pennsylvania. asserted
who → ’re → Pennsylvania
Using advice from Mollick, other experts and tech-savvy friends, as well as tips from AI companies, I set out to test some theories about writing better prompts for travel planning. asserted
I → use → planning
I concocted two fictional five-day trips tailored to the two cities I know best: São Paulo, Brazil, and New York. asserted
I → concoct → cities
I planned to prompt three bots — ChatGPT, Gemini and Claude — trying both the free and paid versions. asserted
I → plan → versions
(The paid versions tended to go deeper and work faster, and were less likely to run out of steam.) asserted
versions → tend → steam
One twist: I went undercover, using new accounts or “temporary chats” so the bots wouldn’t know that I had lived in the two cities for years. asserted
I → go → years
I asked AI to advise a first-time traveller to São Paulo and a veteran New York City visitor who had already seen all the main sights. asserted
who → ask → sights
I was impressed by the results, if not persuaded to give up my old-fashioned predilections for tips from friends, paper guidebooks and late nights of online research. asserted
I → persuade → research
Here are some strategies that worked for me. asserted
that → work → me
Pile on the details Consider this typical rookie mistake: asking a bot to “plan a three-day weekend in Chicago for someone who likes museums and good food.” asserted
who → pile → museums
Experts told me again and again, “Ask generic questions, and you’ll get generic answers.” asserted
you → tell → answers
The results end up being bland — who doesn’t like good food? — as well as vulnerable to “generative engine optimization,” a fast-growing practice in which companies scramble to game the models so they appear in responses. asserted
they → end → responses
(It’s the AI equivalent of search engine optimization, which companies have used for years to rise to the top of Google searches.) asserted
companies → ’ → searches
That includes your specific tastes in travel, your goals for this trip and any advice you’ve already culled from friends or other sources. asserted
you → include → friends
Try compiling a personal travel profile with a few paragraphs about your preferences and philosophy, perhaps in a Google Doc, that you can feed into bots. asserted
you → compile → bots
Mine included my love of “tours with very quirky local guides,” my aversion to international hotel chains and my tendency to choose restaurants based in part on online user reviews written in the language of either the destination (Paris restaurants reviewed in French) or the cuisine (Chinatown restaurants reviewed in Chinese). asserted
Mine → include → Chinese
Mollick urged travellers to be honest, even blunt. asserted
Mollick → urge → travellers
No chatbot will judge you if you say your ideal Chicago trip favours hot dogs over haute cuisine (and neither will most Chicagoans, as long as you lay off the ketchup). asserted
you → judge → ketchup
In addition to your travel profile, you should feed the bot any other research you’ve done using your favourite sources of travel information or tips you’ve gotten from friends. asserted
you → feed → friends
The assistant will work those tips into your itinerary, or perhaps tell you why they don’t fit. asserted
they → work → you
For my São Paulo test, for example, I uploaded advice from a “friend I trust who lives in São Paulo.” asserted
who → upload → Paulo
(It was actually a document I shared with out-of-town guests who came to my wedding last year.) asserted
who → share → wedding
Ask the bot to help AI can actually help you make your prompts more effective. asserted
prompts → ask → bot
Every step along the way, ask the bot if there is anything you can do to make your request clearer. asserted
request → ask → bot
For example, when I uploaded my travel profile, I told each bot to ask me five follow-up questions. asserted
I → upload → questions
For the New York trip, Gemini asked about my willingness to spend several hours travelling to a more distant but interesting part of the city. asserted
Gemini → ask → city
I said that sounded good, but just for one day. asserted
that → say → day
It suggested a trip to Pelham Bay Park and City Island in the Bronx — good advice if I am the target. uncertain
I → suggest → Bronx
ChatGPT suggested Jackson Heights, Queens, saying it “feels almost mandatory for you.” uncertain
it → suggest → you
That made sense, because it also happens to be the neighbourhood I lived in for more than 15 years. asserted
I → make → years
For São Paulo, instead of a day-by-day agenda, I asked for four separate lists — activities, restaurants, nightlife and lodging — with “too many” choices for each one. asserted
I → ask → one
I assumed I could pare down each list by asking the bots further questions. uncertain
I → assume → questions
That worked great, and at the end I asked to group the results into geographically sensible clusters. asserted
I → work → clusters
Know what’s missing Unlike you and me, AI does not have free rein on the internet. asserted
AI → know → internet
Many websites (including The New York Times, which has sued OpenAI, the maker of ChatGPT, and others for copyright infringement) block bots’ access to some or all of their content. asserted
which → include → content
Others have deals with specific AI companies. asserted
Others → have → companies
For example, ChatGPT and Gemini, which is owned by Google, have full access to Reddit — a great source for travel tips — while Claude, which was created by Anthropic, does not. asserted
which → own → Anthropic
…and 27 more, not listed.
💬Give feedback
🕘History 🎫Support