If you recall, Expedia announced in early April that it was going to be one of the first travel players to tap into ChatGPT.
Highlights from the press release:
- Expedia launched a beta version of a new in-app travel planning experience powered by ChatGPT, allowing members to engage in open-ended conversations about travel plans.
- The AI-enhanced travel planning system provides recommendations on places to go, accommodation options, transport details, and activities, based on the user’s input.
- The new system incorporates intelligent shopping, automatically saving hotels discussed in the conversation to a “trip” in the app, facilitating date selection, availability check, and addition of flights, cars or activities.
- Expedia also developed a ChatGPT plugin, enabling users to start a travel conversation on the ChatGPT site and use the Expedia plugin to bring their travel plans to life.
- The AI-based system offers users exclusive discounts and rewards for their trips, accessible directly through the Expedia app.
- Expedia already uses AI and machine learning in its platform to personalize trip options, track flight prices, and provide customer service through an AI-powered virtual agent.
- The collaboration with OpenAI is in beta testing phase and will continually evolve based on user feedback and interaction.
- The new beta experience is being rolled out globally in the English language on the Expedia iOS app.
The Test Trip
To put the Expedia, Bing Chat, Google Bard and Kayak to the test, here’s a simple trip premise I came up with:
- A European family holiday with three young kids
- Unsure of final destination but could be London, Paris or Rome
- Travelling from Washington DC; week-long trip (possibly in July)
- Need to keep overall costs contained
The Expedia App
First, I tried out the Expedia App itself:
Overall, it looks like Expedia has just stuck ChatGPT into their app without any meaningful integration. You end up getting pretty much the same results you would if you were just using ChatGPT separately. There is one little feature that adds any hotel recommendations to a draft ‘trip’ but you’ve got to go check these out separately and there’s no way to get intelligent recommendations based on the criteria specified. Ideally, given the in-app integration, I would have loved to see something more useful, i.e. actually tweaking recommendations based on availability, pricing and proximity to attractions.
The ChatGPT Expedia Plugin
Next, I tried out the ChatGPT plugin Expedia has created to see if it would yield more meaningful results:
This interaction felt a bit better, but overall not much smarter. The ChatGPT Expedia plugin basically pulls in the top 3 hotel recommendations from each city and then “guesses” which would be cheapest. When asked for which dates would be cheapest in a given month, the options that come back are then completely different to the original recommendations. The chat also stops every now and then, needing to be coaxed back into action…not sure why, but perhaps this is due to the token limits on ChatGPT / GPT-4 itself.
How Does Bing Chat Compare?
I also decided to give Bing Chat a shot with the same prompts, given it uses the GPT 4 model and is connected to the web.
In this case, Bing ended up recommending Paris instead of Rome…and basically offered up answers based on various web search results and deferred to Bing Travel for any pricing or booking information, which was disappointing. I also found the lookup and info based on ‘living costs’ irrelevant given the query was focused on summer family travel.
Does Google Bard Perform Better?
Given Google’s existing travel prowess, I was hopeful it would perform a bit better with Bard…and I wasn’t too far off base:
The chat responses, as usual, were pretty fast and Google Bard recommended Rome for the trip. When probed for timing in July it also recommended the week of 10th July for the best fares and provided a handy total for the trip…which was roughly about $2k. I double checked this on Google Travel separately and found that the pricing was pretty optimistic…with the real flight prices being closer to $4-5k. So overall, some good output from Bard, but not super reliable for actual trip planning.
The Kayak ChatGPT Plugin
Given Kayak also has a plugin live on ChatGPT, I gave this a test run too:
The lookups with Kayak took a while (I’ve sped up portions of the video 20x) but overall, I was pleasantly surprised by the overall trip cost estimates provided by the chat. Using Kayak, ChatGPT was able to provide an overall cost for each destination including flights and hotel stays, which was super helpful to understand which city to pick. Things got a little buggy when I asked for specific hotel recommendations, though it did end up providing a specific hotel name and link.
After the tests above, here are a few conclusions I arrived at:
- Feels like Expedia was keen on capturing PR headlines more than delivering a well-integrated, useful product with ChatGPT / GPT-4. Given how new the tech is though, they can be forgiven…and with more time and development, the AI powered features can only get better. Bing Chat (with GPT-4) wasn’t super useful with its mash up of web results and general quality of output.
- The Kayak ChatGPT plug-in, even though it was a little slow, did a nice job of recapping overall trip costs. Google Bard wins for speed and overall quality of output, though the pricing info was a little too optimistic.
- At this early stage in the game, Generative AI is probably best used for general trip inspiration and ideation. Given another 6-12months, we will probably see some terrific travel use cases and (better) integrations from the largest players in travel, from hotel chains to OTAs and AirBnB. Can’t wait to see what’s around the corner!
What are your thoughts on the above tests? Have you tried these tools yourself…if so, please leave a comment or DM me on Twitter or LinkedIn.
0 comments on “Trip Planning Battle – Expedia, Kayak, ChatGPT, Bing Chat and Google Bard Face Off”