While Google remains the top way travelers prepare for their trips, a growing subset of consumers is turning to artificial intelligence to help them research and plan out their stays.
Specifically, they’re experimenting with AI platforms known as large language models, or LLMs. The most popular of these platforms are ChatGPT, Gemini and Llama. Travelers can ask these LLMs to create a travel itinerary for them, and they can add details to their requests to further narrow their results.
Hotel digital marketers have their eyes on this trip-planning trend and have been making moves to increase the chances of their hotels showing up in these LLM-generated results. And, just as importantly, the information these AI tools generate about their hotels must be accurate.
Everyone is researching for a purchase or a trip, said Dan Olsen, director of digital marketing at Springboard Hospitality. Through the LLMs, the top-of-the-funnel research has been outsourced, so the traffic that hotel websites get is more qualified now.
“The silver lining behind it is that by the time they get to our websites, they’ve done their research,” he said. “They more or less know what they want to see, and they are ready to make a purchase or are at least closer to that point of purchase.”
Strategies to show up
It’s important to know how travelers are using LLMs to do their research, said David Solomito, senior vice president of marketing at Sage Hospitality. They may put in a request for an itinerary, but then after reviewing the results, they may refine their inquiries further.
“They’re not just taking the first results, but they’re saying, ‘Oh, actually, on the third day, that sounds a little hectic. I want something that’s closer to where I’m staying and indoors,’” he said. “And within seconds, you have those recommendations. People are also using it for comparisons, so it’s just like that expert in their pocket.”
Sage has a team of AI subject matter experts at the corporate level that hotel teams can tap into for advice. It's one less thing for the on-property teams to worry about, he said. Along with that team, Sage works with agency partners to help as well.
“In practice, that can be done with offering deep and high-quality content on the website in various ways,” he said. “That’s always been a goal, but I think now it’s more important than ever because when you really dig into what it is that ChatGPT looks for, what signals when they come in and say, ‘Here's a hotel that might work for you,’ it's really having a depth of specific information and high-quality content on their site.”
Something as simple as having a frequently asked questions page on a website is useful, Solomito said. Knowing that travelers often search for a hotel with a pool, a best practice would be to have a FAQ question about whether the hotel has a pool so it’s an authoritative answer coming from a trusted source of information.
“Just that simple tweak like that can be the difference of being one of the recommended hotels or places to stop on an itinerary versus not,” he said.
Having authoritative third-party recommendations or endorsements, such as through the press, is important as well, he said. Having that information on the hotel site is good, but having a positive story or recommendation written about the hotel further improves the odds of coming up in an LLM result.
Schema is another factor to consider, Olsen said, adding that digital marketers should think of schema markup as next-level metadata. It’s essentially micro metadata that tells search engines specifically what a certain webpage is for.
“We’re looking for whether it’s local, whether the schema is focused around functional data or purchase data,” he said. “If they’re down the funnel, this is a purchase page. Making the site up very intentionally via schema is huge.”
Concise copy is useful as AI searches like 40 to 60 words, he said. Bullet points are huge. A fast-loading mobile site is important as well. Chatbots and LLMs don’t render JavaScript, so pre-rendering it for them can push websites to the top of the list.
Most searches still go through Google, but they don’t originate there, Olsen said. People are getting exposed to information elsewhere, so the top of the funnel starts outside of Google, and by the time they reach it, they’re further along the process and more ready to complete the task.
Figuring out the way LLMs work is similar to the early days of Google search engine optimization when digital marketers had to understand where the search engine got its information, said Ben Golson, vice president of digital marketing and ecommerce at McKibbon Hospitality. Roughly 30% to 40% of the information that LLMs process, generally, comes from Wikipedia.
That’s led Golson to see if any of McKibbon’s hotels on Wikipedia have historically relevant or otherwise beneficial information on them, he said. McKibbon's hotels also look at online guest feedback.
“It’s really not directly interfacing with the LLMs right now, because there isn’t a specific way to feed it direct information,” he said, adding that Expedia and TripAdvisor, however, have partnered LLMs and feed them first-party data.
That means McKibbon is working with the influencers, especially if they have those partnerships in place with LLMs, he said.
“We’ve had a pretty big push for our hotel teams and [general managers] to focus on what their online reputation says about their hotel,” he said. “Even from where we are today with Google, it pulls keyword information.”
If enough guests are leaving online reviews about hotels in a certain market, there’s a high chance the hotels will start showing up in search results for that market, he said. While the company can’t influence those reviews directly, the property teams can respond to guest feedback directly.
“Going in and responding and having those conversations online is important,” he said. “It can help shape our reputation, essentially, from an online standpoint.”
It’s important to remember that there’s no one right answer, Solomito said.
“It’s learning every day by definition, AI,” he said. “It changes based on what’s being most useful to the users. I think at the end of the day, it’s about providing quality information and content to people generally, and then hopefully tools like ChatGPT will report that.”
Inaccurate information
While LLMs can fulfill a lot of requests, it’s not a perfect process. Sometimes results come back with wrong information, and there’s only so much digital marketers can do about it.
The process to try to fact-check any information from LLMs is still fairly manual, Olsen said. Content teams run through top queries to see whether their hotels are coming up in the results and what information is being given.
Sometimes LLMs pull information from sites that rank well but have dated information, and that’s where the chain breaks down, he said.
“Accuracy [among LLMs] really largely stems from our websites being accurate, because it's essentially just crawling and pulling,” Olsen said. “So, it’s really first and foremost about our sites being accurate.”
LLMs can provide itinerary ideas and provide accurate results for things that don’t change often, such as hotel amenities, but they can have trouble with special events, Golson said. For example, the results for finding a hotel room for an upcoming professional sports game might not be exactly right.
“I did that exact search: 'Find me a hotel room within walking distance for the next Tampa Bay Lightning game,'” he said. “It offered me a four-bedroom home in Brandon, which is an hour away, so we’re not quite there yet.”
It’s necessary to understand where LLMs get their information from currently, Golson said. It’s primarily text-based right now.
“Right now, for me, I think the biggest way that we can influence that is focusing on our content, like written content, and then our guest feedback that's being published online,” he said.
Don’t forget the experiences
Another tip is to make sure the experiences that a hotel connects itself to are evergreen, Olsen said. A common occurrence is a destination may have had a popular restaurant, but that restaurant closed during the pandemic, so an LLM may end up directing travelers to a place that no longer exists.
“It’s just making sure that you’re choosing partners for those pages that are evergreen, that match your business model, that really add value as opposed to there’s a McDonald’s down the street,” he said.
Many travelers, Olsen included, like new experiences that are specific to the destination they’re visiting, so it’s key to focus on the local makers in the market, he said.
It’s also useful to mention on a hotel’s website its proximity to different places in the market, Olsen said. For example, for a hotel in Waikiki, Hawaii, it’d be important to say how close it is to Diamond Head.
“You’re really trying to drive unique value because that’s going to surface to the top of Google for anyone coming into that market, maybe someone who has no idea about your brand,” he said.