A year ago, ChatGPT disrupted our lives and processes. With it, countless AI tools quickly followed suit and became more accessible to the general public.

Search engines are now beginning to introduce a plethora of AI-powered search tools. As businesses, we need to be prepared for how users will adopt these new technologies and how they will alter user behaviors.

Popular and trusted companies such as Alphabet and Meta are trying to speed up this adoption curve by investing significant time and money in raising user trust in, awareness of, and adoption of AI tools.

Trust is one of the key friction points in the AI adoption curve.

For AI to be adopted at scale, users need to overcome any preconceived biases or perceptions of AI in general, including how safe they are to use and the reliability and trustworthiness of the given information.

At the end of March, Google’s blog (The Keyword) focused on ways Google can support users in finding their 2024 travel plans. The first method they outlined is to use Search Generative Experience (SGE).

Google has a history of targeting the travel sector with different SERP features and products, such as the Flights and Hotels features. And so, one of my longstanding beliefs is that any “free to use” public AI tools released by Google will likely be rolled out to and targeting the travel sector first.

This targeting hasn’t gone unnoticed, especially as the Digital Markets Act has come into force. The act makes it so that the likes of Google aren’t allowed to show overly preferential treatment for their own products and services in search results.

AI offers a way around this. Google can offer its own solutions and avenues for paid placement while keeping users within the Google ecosystem.

How disruptive could AI be to itinerary research queries?

All the travel companies I’ve worked with rely on organic traffic for every funnel stage, and AI has the potential to impact all phases of the user journey from destination discovery to itinerary planning. Because of this, travel companies stand to lose both direct bookings and sales, on top of missing brand touchpoints during the discovery-focussed phases.

To analyze the potential impact, we’ve taken a sample of 50 destinations across Europe, North America, The Middle East, Africa, Asia, and Oceania and asked four generative AI tools to plan two-day itineraries for the destinations.

The four AI tools we’ve used are:

  • ChatGPT (GPT4, plus browsing)
  • Search Generative Experience (Labs)
  • Copilot
  • Meta AI

By looking at the wider ecosystem, we can ensure a more comprehensive analysis and understand user behavior on both Google’s and other large language models (LLMs).

Methodology

We asked the four LLMs to outline two-day itineraries for 50 cities across Europe, North America, Africa, the Middle East, and Oceania/Asia.

We chose the cities based on their reported popularity for 2024, using articles from The Times, TripAdvisor, and USNews to compile the list of 50.

Our findings

From the data we analyzed across the four LLMs, we found that:

  • SGE recommended the fewest itinerary items across the 50 cities, with 344, and was the only one to not make it into the double-digits for each region.
  • Meta AI recommended the most itinerary items across the 50 cities, with 742.
  • Meta AI was the worst offender regarding hallucinations, by recommending 14 of 742 places that didn’t actually exist (1.9%).
  • Three LLMs produced more itinerary items for Europe than any other region, with the exception of SGE. SGE returned five more items for North America than Europe.
  • All four LLMs produced the fewest itinerary items for Africa than any other region, with the Middle East consistently producing the second fewest.

From a personal perspective, SGE’s restrictive results were surprising given how much Google has invested in its AI technologies, the volume of data Alphabet has access to, and the perceptions of each LLM’s capabilities.

Average Places Mentioned SGE      Copilot Meta AI ChatGPT-4
Europe 7 13 20 13
North America 8 10 19 12
Middle East 7 7 11 10
Asia/Oceania 8 7 11 12
Africa 7 7 13 11

Hallucinated (fake) places

In our test, AI recall was generally hallucination-free.

When using the previous model of ChatGPT, ChatGPT-3, in initial testing, the results were largely inconsistent. The LLM struggled quite a bit more, and results varied greatly — from highly accurate to highly hallucinated.

Focusing purely on ChatGPT-3 as a testing source would be unreliable without repeating the same study enough times to gain statistical significance. Furthermore, this wouldn’t reflect real-world user behavior, and the resulting data would be misleading, with hallucination percentages ranging from 5% to 95%.

“Bad” or inappropriate places

In addition to looking for hallucinations, we reviewed the itineraries for questionable places that didn’t “fit the bill”.

For example, ChatGPT-4 recommended a children’s playground in Berlin for an afternoon stop-off, which, to an AI, seems like a safe choice. But upon reviewing the playground ourselves, it looked fairly run down, with rampant graffiti. Not really the type of place you’d usually add to your itinerary as a traveler on a weekend city break.

That said, the percentage of bad or inappropriate places was relatively stable across all four LLMs. While Meta AI returned the most number of inappropriate places, it also returned the most places overall, keeping its percentage in line with the other LLMs.

How does this impact Travel SEO & online discovery?

AI technologies are poised to greatly disrupt all aspects of the travel industry.

As none of the LLMs have been rolled out to the general public, we’re yet to see the impact or penetration these technologies will have. The only thing of which we can be certain is that there will be an impact.

From our wider testing of LLMs across the travel sector, we found that generative AI could streamline the customer journey in determining destinations, accommodations, attractions, and itinerary planning.

This threatens travel brands and their traditional acquisition of users, sales, and brand visibility. However, it also presents an opportunity for travel brands to be more competitive by learning how to use AI to their advantage.

Search Generative Experience

As I mentioned above, Google has usually first targeted the travel sector with its innovations. The fact that they’re already starting to push the SGE beta to travel signals its impending rollout.

While SGE didn’t return as many itinerary items as the other LLMs in our test, its integration with Google’s other data sources — e.g., Flights, Hotels, and Maps — means it will disrupt customers’ current journeys.

If you’re interested in learning more about SGE and the impact it will have, we’ve covered Google’s AI search products on our YouTube channel:

The Meta AI threat to travel search

In our test, Meta returned more than double the number of itinerary items than Google’s Search Generative Experience. However, it wasn’t 100% accurate and had a 1,8% hallucination rate. (Which, to be fair, is still much lower than early ChatGPT-3 tests.)

Like Google (and its parent company, Alphabet), Meta products already have considerable market penetration. Statista reports Facebook has 3 billion monthly active users (MAU), while Instagram and WhatsApp have 2 billion and Facebook Messenger at 1 billion.

If Meta integrates Meta AI into their core products, the adoption curve of AI will be greatly sped up across both the business-to-business (B2B) and business-to-consumer (B2C) industries. Whatsapp and Messenger products are already being utilized for B2C communications in addition to facilitating payment and money transfers.

Further, several studies (including one from the Pew Research Center) highlight how millennials have historically led the technology adoption curve. Now, Gen Z is leading the adoption of AI technologies. A British Ofcom study found that 40% of 7 to 12-year-olds use AI. However, that nearly doubled for 13 to 17-year-olds, where 79% use AI. Meanwhile, that figure drops considerably when considering anyone over the age of 16: just 31% of adults (16+) have adopted AI technologies.

As the user base of Meta’s core products falls within the 18 to 35 range, focusing on improving the adoption rate for AI makes a lot of sense for the tech giant.

Preparing your travel SEO strategy

Travel brands that don’t prepare and adapt their strategies to accommodate the impending AI disruption caused by various LLMs will experience an impact on their traffic and overall brand visibility. This will inevitably affect their bottom line.

Although you can’t directly optimize for AI, your strategy must include a defensive approach to it.

By understanding your exposure and the potential impact on discoverability, you can develop tactics to grow relevant traffic and reach your audience through other queries and touchpoints.

If you found this article useful, you should also read our thoughts on how the travel sector can better utilize AI for data and SEO. Alternatively, if you want to develop your AI travel SEO strategy, please get in touch.