Cambodia Tourism Technology in the AI Itinerary Era
- Camden
- 23 hours ago
- 5 min read
Cambodia's tourism sector generated 3.87 billion dollars in gross revenue in 2025, up 6.6 percent year on year, while supporting roughly 420,000 direct jobs. This morning I was at the Ministry of Tourism in Phnom Penh with Casey, a friend and colleague from Malaysia who built an AI-driven tourism application in that market and grew it from about 30,000 active users to more than 400,000 in a short period. It was my pleasure to make the introduction. The question we brought into the room was not how many more visitors Cambodia can attract. It was how much of what those visitors spend actually reaches the merchant. Cambodia tourism technology is where that answer gets decided.
A user curve worth reading closely
The figure that matters in Casey's numbers is not the total. It is the word active.
Follower counts measure attention. Active users measure behavior, and behavior is the only input that converts into revenue at a counter. Applications that reach that kind of adoption in a tourism market almost never win on marketing. They win because they solve a coordination problem that both sides of the transaction were solving badly on their own. A traveler wants to know what is nearby, what is open, what is worth the ride, and what it should cost. A small operator wants distribution and has no realistic way to buy it. Software sits between those two positions and closes the gap.
That is a category logic, not a national one. It has produced similar curves across several ASEAN markets, and the underlying conditions in Cambodia are strong.
The planning funnel moved before the booking did
The behavioral shift that makes this urgent is happening at the front of the funnel rather than the back.
A Klook survey of 11,000 travelers found that 91 percent now rely on AI travel planners at some stage of a trip. Skift and McKinsey recorded a 124 percent year-on-year increase in travelers using AI tools extensively for planning. Booking behavior has moved more slowly. Close to 70 percent of travelers still complete the transaction through a brand they already trust, and OpenAI withdrew its in-chat booking function in March 2026 after finding that users would research in the chat window and then pay somewhere familiar.
Read those two facts together and the conclusion is specific. The itinerary is increasingly written by a model. The payment is still made to a merchant. Between those two moments sits every business that will either appear in the generated answer or will not exist as far as the traveler is concerned.
A listing on a booking site was a shelf. A model output is a shortlist. Shelves show inventory. Shortlists are short.
Where Cambodia tourism technology actually earns its return
The payment layer here is already built, and built well. The National Bank of Cambodia launched Bakong in 2020, the KHQR standard gave the country a single interoperable code across banks and wallets, and a dedicated Bakong offering for international visitors has been in development to let travelers transact without a local bank account. A tourist can pay a tuk-tuk driver, a guesthouse, and a noodle stall through the same rail.
What is less developed is the layer above the payment. Discovery, language, pricing, availability, booking, and review response are still handled manually by the operators who capture the largest share of visitor spending in daily terms.
The official inventory gives a sense of scale. Cambodia counted 95,588 hotel and guesthouse rooms, 2,301 licensed restaurants, and 652 licensed tour operators and travel agencies. That formal layer sits on top of a far larger population of family-run businesses that serve visitors every day without appearing in any registry. Alongside international arrivals, domestic tourism reached 25.16 million travelers, an increase of 11.7 percent. Domestic demand is the volume base most foreign investors never model.
Tourism revenue does not arrive at a country. It arrives at a counter.
The small operator is the product
The version of AI adoption that fails in emerging tourism markets is the one that asks a guesthouse owner in Kampot or a restaurant family in Siem Reap to become a technologist. The version that works removes work. Menu translation into four languages. A price that adjusts to the week rather than the season. Inventory that updates without a phone call. A review answered in the correct language within an hour instead of never.
The structural piece Casey described is access. A small hotel, a tour operator, or a family restaurant that cannot fund a website and has no one to maintain it is not going to build one. Inside a shared platform, the same operator can build and update its own presence directly, at no meaningful cost, and be legible to the systems travelers now consult first. The economics of the content layer are handled the same way. Creators and influencers earn from advertising revenue generated inside the platform, which gives the storytelling side of the product its own reason to keep producing rather than depending on a marketing budget.
That opens the contribution layer to residents and foreign long-stayers as well as businesses. Routes, neighborhoods, seasonal detail, the small operators nobody has written up. Cambodia is known through a short list of famous sites, and there is a great deal more of it. A platform that lets thousands of local operators and residents write their own entries adds far more of the country to the material travelers are planning from.
A guesthouse in Kampot without a website does not lose a booking. It loses the chance to be mentioned.
Why a property desk pays attention to this
Tourism sits beside construction and real estate as one of the four pillars of the national economy, and the two pillars are linked more tightly than most investors price.
Visitor spending underwrites ground-floor commercial rent. It underwrites serviced apartment occupancy, short-stay yield in central districts, and the viability of retail podiums that developers build on assumption rather than evidence. It determines whether a provincial node with a new road and a new airport becomes a market or stays a map reference. When spending per visitor rises, commercial rent follows it with a lag. When the merchant layer digitizes, that spending becomes measurable, and measurable revenue is what banks lend against and what buyers underwrite.
A country that can demonstrate transaction-level tourism data is a country whose commercial property can be valued on cash flow rather than sentiment. That is a slow change and a structural one.
The conversation ahead
Casey and I hold a straightforward view of this country. We like it here, and we would like to be useful to it. Today was a first conversation and nothing more, an introduction between people who have built something that worked in one ASEAN market and a ministry responsible for a sector carrying a meaningful share of national output. What comes next is technical work, and technical work is best done quietly.
The demand signal is not in question. Traditional businesses are being asked to operate with technology whether or not anyone builds the tools for them, and the operators who get those tools first will take share from the ones who wait. That pattern has repeated in every market where the payment rail arrived before the software layer.
Cambodia's tourism question is no longer volume. It is capture, and capture now begins in a chat window rather than a search bar.
Investors who track arrival counts are reading the headline. Investors who track spend per visitor, and the merchant infrastructure that captures it, are reading the balance sheet. The second number moves commercial rent and short-stay yield long before it appears in a press release.
At My First Corner, we follow tourism data because it prices ground floors, serviced units, and provincial commercial stock before the repricing shows up in asking rents. The conversation is available when it is useful.
Comments