Travelers are increasingly turning to ChatGPT's generative AI engine to conduct flight searches. And as they do, OTAs and travel metasearch have an advantage over airlines.
In September, ChatGPT accounted for 2.3% of the referral traffic airline websites received from search engines, according to an analysis by PROS, an airline retailing and revenue management company that works with more than 150 carriers. While that figure is relatively small, it's growing fast: ChatGPT generated just 1.8% of referral traffic for airlines in August, and it is up more than 230-fold since last November, the first month OpenAI offered ChatGPT Search.
ChatGPT is currently the dominant force in GenAI search for air bookings, accounting for 95% of airlines' AI engine referral traffic, according to PROS. That number is poised to be challenged as other technology companies develop and refine their AI search engines.
Within ChatGPT, however, airlines are getting short shrift. In an analysis of referral links within the conversational responses ChatGPT provides to flight queries, PROS found that by a wide margin they favored OTAs and metasearch sites over airline websites, thereby sending extra traffic to aggregators such as Kayak and Expedia instead of to airlines.
Specifically, the analysis found that 41.8% of brands mentioned in flight-related conversations by ChatGPT didn't get the associated link. Among airlines, the unlinked rate is 74.6%, with referrals instead going mostly to an array of OTAs and metasearch engines but also 5% of the time to Wikipedia.
The unlinked rates for OTAs and metasearch sites were just 8.8% and 8.6%, respectively.
The results speak to an emerging problem for airlines if they hope to avoid losing share to aggregators, said Enmanuel Tirado, PROS head of search engine optimization (SEO) and digital analytics.
And, said Vik Krishnan, an aviation-focused partner for the consulting firm McKinsey & Co., it's an example of the many unintended consequences that can be expected in the flight and travel booking spheres as the conversational responses of AI search engines encroach upon or overtake the straightforward links that search engines have historically produced.
OTAs and travel metasearch sites, though currently in an advantaged position on ChatGPT search, also face threats.
Already, said Krishnan, many websites that rely on click-through traffic as a source of demand have seen 20% to 30% drops in visibility as Google and other traditional search engines have begun responding to queries with AI overviews positioned above the usual link-format responses and as consumers shift to AI search engines.
That's a threat to OTAs and travel metasearch engines that rely on Google ads and SEO to generate traffic.
It's also a threat to travel suppliers, said Krishnan, such as a Mexican hotel that pays for advertising links in Google search query responses about Mexican vacations.
Google itself isn't immune, either. The Silicon Valley behemoth is devising strategies to replace lost SEO advertising money, including paid advertising within its AI overviews.
"The story on how AI engines are going to monetize eyeballs all the way through to the transaction -- that script is still being written," Krishnan said.
One reason ChatGPT favors aggregators, Tirado said, is that they are rich in fare and flight information and include flight comparisons. Airlines could improve the link imbalance by making sure they have a separate webpage for each flight with fare details across cabins and products and complete with flight durations and schedules.
Airlines should also alter their technology stacks to accommodate ChatGPT, including making sure ChatGPT bots can access their webpages, Tirado said.
Mike Coletta, senior manager of research and innovation at Phocuswright, said one thing AI search engines aren't yet good at is finding the most up-to-date airfares and hotel prices, which are contained within GDSs, OTAs and airline websites.
But a technology standard called Model Context Protocol (MCP), released in April by AI research company Anthropic, enables AI search engines to obtain that data by sending structured queries to travel suppliers and aggregators.
Turkish Airlines, online travel agency Kiwi and Sabre are among the travel companies that thus far have built out MCP servers to allow access to AI engines.
As airlines and travel aggregators fight for the attention of emerging AI search engines, being an MCP leader could be an advantage.
"It's looking like a good possibility that every travel company will need an MCP server," Coletta said.