Artificial Intelligence and Trip Planning

Adventure is a State of Mind

Artificial Intelligence

Artificial Intelligence (AI) is definitely here to stay.

I occasionally experience brief fears of a dystopian society in which AI rules our world. (I think I’ve watched too many science fiction movies.) Most of the time, though, it doesn’t have a direct impact on my daily life (yet). In fact, I’ll admit that I wasn’t even paying much attention to the news — when the chatbots made a significant leap in development and ChatGPT was introduced — until some of my friends who are instructors at the local university began talking about how the new level AI will impact their students’ writing, and how they will need to change how they assess and grade their students.

Then someone mentioned AI building trip itineraries. That got my attention, of course.

Travel and AI

Is AI the future of travel planning? In a few seconds, with the help of AI, can we all plan the perfect trip? Do we want to?

Human beings are inherently lazy creatures, hence the constant search for the most efficient way to do everything. Almost every technological advance has been an attempt to make something easier for someone. As a human, my instinctive reaction to the idea of AI planning the perfect trip is… YES! Please!

I love planning and developing trip itineraries, but I’m all for letting AI do it instead if it can be done for me quickly, inexpensively, and created with my specific requests and parameters. (My reason for choosing independent travel over organized tour packages is that organized tours limit where I can go and what I can do.) However, you know what they say about something that sounds too good to be true….

So, I decided to do a little research and share my results and thoughts. I’ll tell you up front that my research did not cover all of the possible options in the AI world. There are a lot of options! And it is such a fast-growing field that between the time I post this and you read it, there will probably be even more AI websites and apps entering the marketplace.

The following are my thoughts based on my personal experiences with each. I’m not promoting any over another, and I recommend that you conduct your own research/experiments with AI and form your own opinions. That being said, at the end of this post, I’ll share with you some of my overall thoughts regarding the future of travel and AI.

The Research Subjects

I tried to sample some of the more well-known generative language AI tools (ChatGPT, Bard, and Copilot) as well as some of the travel-specific tools (Trip Planner AI and Roam Around).

• ChatGPT – The company OpenAI created ChatGPT and launched it in November 2022. It is a form of AI that uses language processing to create human-like conversations, including answering questions. ChatGPT also is designed to constantly refine its answers. In human terms, it’s learning and improving all the time.

• Bard – Google, which created Bard, defines it as a “conversational AI tool.” It’s similar to ChatGPT in that it is able to provide answers to questions and summarize information in a conversational manner.

• Copilot – Copilot is Microsoft’s contribution to AI and is also sometimes called Bing Chat. It is also similar to ChatGPT and Bard in how it provides a conversational experience when providing information or summaries.

• Trip Planner AI – Trip Planner AI is a website using artificial intelligence to create customizable travel itineraries.

• Roam Around – Roam Around’s promotional materials describe themselves as incorporating ChatGPT into their system to provide millions of prepared itineraries for different travel destinations.

My Question/Prompt

Greg and I are in the very initial stages of planning a trip to Greece, meaning we know who we want to travel with, but we don’t have dates yet and have only a vague idea of where we want to go. I decided this was a perfect opportunity to try out some AI options to see how they can help.

For the AI systems that have an open communication option – you type in a prompt or a question – I used the phrase, “Plan a week in Crete, Greece, staying near Chania.” This worked for ChatGPT, Bard, and Copilot.

For Trip Planner AI, I filled in the boxes or checked choices as requested, providing information such as location, how busy I want the trip to be, and budget. For Roam Around, initially you can only give a location and dates, although later there was an opportunity to give more information about what you want to do on each specific day.

The Results With the Open Communication Options

I began with the open language systems. For all of them, I was amazed at how quickly the responses began scrolling down my laptop screen. They all began with fairly flowery prose about the “beautiful island” with a “rich history” and “stunning” landscapes and beaches, which shouldn’t have surprised me as much as it did given they are probably acquiring a lot of their information from travel websites that use exactly that language.

Each AI system gave me an itinerary with multiple activities in a day. AI definitely does not want anyone to be bored! They all appeared to group the activities geographically, although I didn’t check every single activity, and it is possible that there are activities grouped in one day that are actually on opposite sides of the island. None of them provided how much time it takes to travel between locations, although Copilot did provide distances from Chania for many of their recommendations.

There were many similarities, which didn’t surprise me since I assume all of the AI programs are able to detect popular activities and destinations. For example, each of the three options recommended taking a day trip to the Samaria Gorge, which ChatGPT and Bard both pointed out is the longest gorge in Europe.

Interestingly (and maybe a little alarming), Copilot did not take arrival and departure into account and filled each day, including the first and last days, with a long list of activities. Both ChatGPT and Bard referred to arriving, picking up a rental car, and taking time to unpack and settle in on the first day, as well as recognizing that the last day would not be a full day due to departing the island.

ChatGPT gave the briefest responses and did not discuss food other than recommending eating traditional food at a taverna (a small Greek restaurant). Bard did not recommend any specific restaurants but did discuss local food specialities to try. Copilot didn’t mention food at all. (Quite disappointing, given that one of my favorite things to do when I travel is to try the local food.)

Overall, Bard gave the most description with its recommendations. And a bonus from Bard was it provided lots of photos.

Just to check out what sort of extra information each of the AI platforms can provide, I asked each of them an additional question – “What should we eat in Crete?” (Of course I did!) They all provided responses, but there were significant differences in the quality of their suggestions. Copilot just provided two recommendations, but they were unique to Crete and not just general Greek recommendations. ChatGPT gave 15 food and drink suggestions, but they were all general Greek foods. (Yes, almost everyone, even those who have never been to Greece, knows that baklava is a delicious Greek dessert.) This is where Bard really outshined its competitors. It provided me with 12 food suggestions, several of which are Cretan specialties, along with a photo of each food!

The Results With the Travel AI Options

I have to admit that after an overall positive experience with the general language-based AI options, I had high hopes that the options focusing specifically on travel and itinerary creation would provide even more useful information. Unfortunately, although there were some benefits to using the travel specific platforms, I’m not sure that the benefits outweighed the limits.

Both Trip Planner AI and Roam Around were fairly limited as to what can be asked. Both of them had a prompt box to put in a location (as compared to being able to ask an individualized question), but Trip Planner AI only allows you to put in a city name rather than a country.

There was an option to add additional cities, although it did not appear to consider any travel time between the two destinations when a city was added to the itinerary. (I added Athens to my Chania request, and it gave a full list of activities for both the last day in Chania and the first day in Athens.) Roam Around also gave a later option to add a city, although when I attempted to add Athens to my Chania request, it changed the title of my itinerary but did not actually give any different or additional information.

Trip Planner AI had some extra features as far as being able to give it information about how many people you plan to travel with, the budget, and how busy you want to be – Chill, Touristy, or Packed were the options. Roam Around, on the other hand, had a prepackaged, one-size-fits-all feel to its results. It reminded me of looking through the promotional materials of a major tour company.

The most helpful aspects of both of the itinerary-creating AI programs that I tried was that they provided some extra information regarding specific locations. Roam Around provided links to recommended restaurants and where to buy tickets for some activities and historical sites. Trip Planner AI also provided specific restaurant recommendations with computer links to information and maps, as well as links to the websites of activities, and prices. Perhaps most helpful, it gave estimated times to travel between recommended activities.

Will I Use AI in the Future?

I definitely think AI is another tool we can all use in planning trips, but it’s not capable at this time of doing all of the work required in planning a travel adventure. From what I’ve seen, it’s the equivalent of talking to a friend about the trip they took to someplace you’re interested in visiting. If you’re not sure what’s in the area, AI can give you some general ideas and descriptions of the options. If you ask the right questions, it might even be able to narrow down the activities to specific ones that would be the most appealing to you, much like a friend who knows you really, really well might be able to do.

I think it would be a big mistake to accept AI-generated itineraries without doing additional research, especially regarding time spent traveling between locations. And ultimately, you will need to make the final decision as to how many things you want to do in a day and which activities you want to prioritize.

For me, AI might be a good starting point when I don’t know much about an area. I’ve previously done this by looking at suggested itineraries in guidebooks or on websites, and this provided a very similar experience. I think in the future we will be able to ask more specific questions of AI, such as asking for an itinerary that takes into account our specific interests and abilities, and we will then likely get a more personalized itinerary than a guidebook could ever provide. (Maybe AI will be reading our minds by then, but I was thinking more along the lines of AI having access to our internet history and knowing what we tend to spend our money and time on.)

I could have asked for specific information about prices and travel distances in the general open communication AI programs, but that’s not really any different than doing an internet search that we’re all used to doing. Perhaps asking a better question, such as requesting a one-week itinerary with travel times included, would fine tune the AI responses somewhat.

In Summary

I don’t think travel agents and tour guides will be completely replaced by AI anytime soon. I also think that independent travel adventures will still require significant time and planning by those of us who wish to create our own tailor-made trips.

In a previous post, I discussed how knowing how to read reviews can make a difference in how useful they are to you. Similarly, it will be important to read AI-generated information with a critical eye rather than accepting it completely without any analysis. The quality of the information you can get from AI will largely depend on how you word your questions and requests, but it will also be important to read results with discernment, rather than naively accepting all of the information as accurate.

AI models have been known to “hallucinate” and produce inaccurate or misleading results, so it is important to double-check the responses you get. In other words, use a little common sense. In the Crete examples, common sense suggests that you are unlikely to be able to participate in a long list of activities on the days that you are scheduled to arrive and depart, regardless of what AI is recommending.

Reading furnishes the mind only with materials of knowledge; it is thinking that makes what we read ours.

John Locke

AI seems to be a helpful tool, however, especially at the beginning of the planning stage. For example, if you want to travel somewhere you’ve never been, AI can be helpful in giving you highlights of the area, narrowing down options, and discovering what’s available in an area. This is something I’ve previously done by reading and researching as much as possible about an area when I initially identify a travel goal. It’s how Greg and I decided we want to stay in or near Chania when we visit Crete, even though we’ve never been to Crete before and didn’t know anything about any of its cities until we started researching it.

Interestingly, I decided to see whether AI might come to a similar conclusion as I did from my own research. I asked Bard, “Where should I stay in Crete if I’m interested in history, nature, beaches, and good food?” Bard’s top recommendation for me? Chania. Hmm, maybe AI is going to play a big role in my travel planning in the future.

Have you considered using AI as a tool in planning your next travel adventure? How do you think AI can best help you?