OpenAI’s decision to scale back plans for direct travel bookings within ChatGPT and direct transactions to third-party apps highlights the complexity of the booking process, adding another layer to agents’ ongoing battle with online travel agencies (OTAs).
Earlier this year, OpenAI announced that it would not go ahead with plans for direct bookings and would instead route bookings to OTAs such as Booking.com and Expedia. Following the announcement, shares at both companies rose by around 10%.
“OpenAI recognises the complexity of travel transactions it hadn’t fully understood before. Most forms of travel, particularly hotels and flights, have an entrenched value chain with multiple players involved in standard transactions. Add to that the complexity of on-demand inventory and slim margins, particularly on flights,” said Brennen Bliss, CEO of Propellic.
Otto de Vries, CEO of Asata, said there was also a financial perspective. “AI platforms are under growing commercial pressure to move beyond subscription revenue, and routing bookings through third-party OTAs is one of the cleaner ways to monetise at scale. The fact that OTA share prices jumped on the news tells you that the market believes established intermediaries are not going away,” said De Vries.
Threat to agents
AI pushing transactions through global OTAs is not a new threat as Google has operated a similar model for years. “We’ve operated in an algorithm-driven environment for years. The difference now is the technology behind it,” said De Vries.
This could have an impact on smaller agencies with limited resources. “OTAs have the inventory depth, the conversion infrastructure, and now the preferred position in AI-driven booking flows. Smaller agencies do not always have the resources to integrate into these platforms in the same way. AI accelerates a consolidation that was already under way, pushing more volume toward large platforms and making it harder for independent agents to compete on visibility alone,” said De Vries.
However, Bliss says visibility is dependent on the model customers use. “For instance, Google’s AI mode gives preference to operators, whereas ChatGPT seems to give some preference to the OTAs, but the jury is still out and I'm sure that will change fairly frequently over time.”
Managing complexity
AI and OTAs still lack the capabilities of managing complexity. “A self-booking tool works reasonably well for a straightforward return flight and hotel booking. It falls apart the moment something goes wrong or when the trip is complicated,” said De Vries.
This is particularly relevant in South Africa, where most outbound travel involves long-haul routes, complex connections, visa requirements and health documentation. “These are the trips where the cost of something going wrong is highest, and where travel consultant expertise has the clearest commercial justification,” said De Vries.
Instead, AI is better positioned as a discovery and inspiration tool. “AI is good at helping a person figure out what travel products they should be purchasing and what would align with their goals, as it is able to personalise much better than using a search engine. It can be about that specific person, their family, their situation and their budget,” said Bliss.
Data responsibilities
As travel bookings become fragmented across AI platforms and OTAs, concerns are emerging around data security and accountability, particularly after Booking.com recently experienced a global data breach.
“When AI platforms route consumers toward third-party OTAs, the question of who is responsible for data security and who the consumer holds accountable when something goes wrong becomes unclear. Under POPIA, South African businesses handling personal data carry real obligations. How those obligations are distributed when an AI platform, an OTA, and a consumer are all part of the same transaction is not yet settled,” said De Vries.
