For Chinese travellers, AI now forms the list of destinations before any enquiry reaches a desk, and it favours easy visas, short flights, a strong safety record and plenty of content in Chinese. For inbound operators in Egypt and the Gulf, content in Chinese that a model can read is now part of distribution.
Key facts
- 240 travel questions were put to five major Chinese AI platforms, using eight traveller profiles — Create Consulting China study, reported by Skift.
- Four things the platforms leaned on: easy visa access, short flights, a strong safety record and plenty of content in Chinese — the same study, via Skift.
- More than 4,400 partner companies are documented by name across the desks the author ran — the author's operations archive.
- More than 410 hotels were chased directly in one operations mailbox — the author's operations archive.
The news. Create Consulting China put 240 travel questions to five major Chinese AI platforms, using eight traveller profiles, Skift reports. Broad questions returned a small set of familiar, nearby, visa-friendly destinations; places such as South Africa and Mexico appeared only when the question named a continent.
Is the shortlist made before the enquiry?
Skift sums it up exactly: "Travel marketers have spent decades convincing Chinese travelers to pick them. Now they have to convince AI to mention them."
The study's point is where the influence sits. Not in choosing between two hotels, but earlier, when the list of destinations is formed. A destination that misses that list does not lose the booking. It never enters the conversation. No desk, no DMC and no sales trip reaches a traveller who never heard the name.
Why read the filters as a checklist?
The platforms leaned on four things: easy visa access, short flights, a strong safety record, and plenty of content in Chinese. Egypt and the Gulf both court Chinese travellers, so this is close to a checklist.
Visas and flight distance are policy and geography. No travel company moves those. Content is different. Pages in the traveller's language that a model can read are something a tourism board, a hotel group or an inbound operator can produce now. The study also found the models are more conservative than the travellers, and that compounds: destinations already mentioned get mentioned again.
Phrasing mattered as well. The same intent, asked two ways, made destinations appear or vanish. That is why I publish the numbers from my own archive openly, with the method beside them. A model can repeat a fact it can find. It cannot repeat a brochure.
Where does the desk still stand?
AI builds the shortlist. On Skift's reading, travellers still turn to booking platforms to compare and pay. Between those two steps sits the part no shortlist sees: the operator who knows which hotel is really ready for the group, and the partner who answers on a public holiday.
Across the desks I ran, more than 4,400 partner companies are documented by name, and in one operations mailbox alone more than 410 hotels were chased directly, one property at a time. That network appears in no AI answer. It only gets used if the destination made the list first.
Why it matters to the agency: for inbound operators in Egypt and the Gulf, content in Chinese that a model can read is now part of distribution. Without it, the best ground handling in the region is never asked for.
Source: Skift — The Race for Chinese Travel Demand Is Now a Fight to Get on the AI Shortlist.