Kakao Expands AI Commerce Inside KakaoTalk: Why Payment, Not Recommendations, Is the Real Test for Kanana

The Day KakaoTalk Becomes the Starting Point for Transactions

Kakao expands AI commerce tests; payment is key beyond recommendations

Kakao is expanding AI commerce experiments inside KakaoTalk, aiming to keep users within the app from recommendation and search to booking and payment.

www.digitaltoday.co.kr

KakaoTalk, the messaging app used by 50 million people a month in South Korea, is trying to become a starting point for transactions rather than a simple way to stay in touch. Kakao has rolled out features in quick succession: AI suggests who to send a gift to on the first screen of the Friends tab, finds places that match the context of a conversation, and links users through to reservations.

The aim is simple. Pick up demand that arises inside a KakaoTalk conversation before the user opens a search window, and complete the booking and payment inside the app. If users finish shopping in KakaoTalk before moving to an external search or e-commerce app, Kakao can capture not only ad exposure but also brokerage fees, payment fees and an increase in commerce gross merchandise value (GMV) all at once.

Still, the original report keeps returning to one point: the real prize is payment integration that goes beyond recommendation. Simply having AI suggest people to gift or places to visit has limits for monetization.

Two Mechanisms That Turn Conversation Context Into Transactions

Kakao recently added Special Friend to the top of the KakaoTalk Friends tab. AI recommends suitable gift recipients by considering anniversaries, friend relationships and usage history. It also reads gift demand from conversation context. If a user sends a friend "Good luck on your exam," for example, it surfaces gift recommendations tied to exam support.

The second pillar is a stronger place agent inside Kanana, the AI service within KakaoTalk. Beyond restaurants, it now recommends tourist sites, accommodation, exhibitions, performances, cinemas, gas stations, convenience stores and auto repair shops based on conversation context, and some can be booked directly within KakaoTalk. If a user registers a visit schedule for a specific area, a briefing also provides local restaurants, parking and menus.

What both features share is a design that starts from data a general shopping mall struggles to have: a friend's birthday, exam support, a travel schedule. The report frames conversations, schedules, relationships and gift history as the distinctive competitive asset of Kakao's commerce.

Solid Earnings, but the Market Is Asking About Payment Conversion

The commerce business results themselves are relatively solid. Kakao's 2024 integrated commerce GMV reached 10.6 trillion won, up 6 percent year on year, and commerce revenue rose 7.8 percent to 951.0 billion won, accounting for 11.7 percent of total revenue. According to consensus from financial information firm FnGuide, second-quarter 2026 revenue is estimated at 2.0492 trillion won and operating profit at 226.3 billion won, with operating profit forecast to grow 21.7 percent year on year.

That profit improvement, however, needs to be viewed separately from the AI business itself. The main driver of rising operating profit is cost cuts from streamlining affiliates, whose number fell from 147 in May 2023 to 94 at the end of last year. Even without a sharp rise in revenue, costs fell, so operating profit increases relatively more.

The problem is that these results are not persuading the market. Ho-yoon Jung, an analyst at Korea Investment & Securities, notes that the mainstream view has become to treat the internet industry as low-growth domestic stocks as growth slows. Eun-jung Shin, an analyst at DB Financial Investment, adds that cumulative users of ChatGPT for Kakao are estimated at about 11 million, but traffic indicators and external partner connections have not yet been activated. Both suggest Kakao must show that agentic AI can lead to actual revenue.

Even With OpenAI, the Wall of "All the Way to Payment" Remains

Kakao has already built some experience linking external partners. ChatGPT for Kakao, a chatbot-style service co-developed with OpenAI, has completed links via KakaoTools with Olive Young, Musinsa and Saramin. But it remains at the level of conversational information search and lookup, and end-to-end commerce that extends to payment is still a task to be proven in Kanana in KakaoTalk.

Brokerages are split on the outlook. Dong-hwan Oh of Samsung Securities says both the Kanana and ChatGPT tracks are delivering weak results and calls for a redesign from the consumer's perspective. Chang-young Lee of Yuanta Securities reads the same situation differently, emphasizing that Kakao already has shopping, maps, pay, mobility, music and webtoons plus about 50 million monthly active users, so it can enjoy a first-mover advantage from the most favorable position. Views are split between those pointing to a lack of completeness and those focusing on existing assets.

Seok-oh Kang of Shinhan Investment Corp expects Kakao to unveil an agent in early August that can handle search, recommendation and payment within KakaoTalk. The key is whether it can bring apparel, cosmetics and travel into KakaoTalk commerce beyond existing gifting. That said, the official launch timing and the level of fees remain undisclosed, and this expectation is only an analyst's projection.

Naver Runs Ahead, and the Battle for Korea's Commerce

Kakao's move is not a solo experiment but part of a battle for leadership among Korea's platform giants. Rival Naver launched an AI shopping agent in February 2026 that takes natural-language requests, searches across e-commerce platforms and completes checkout. The two strategies contrast sharply: Naver takes a closed path that tightly integrates its own shopping, maps and blog, while Kakao takes an open path, connecting external AIs to roughly 200 services through its MCP-based platform PlayMCP.

Payment integration sits at the core of this rivalry. If actual payment after a recommendation happens in an external app, the revenue Kakao can secure is limited. Conversely, if it can keep everything from conversation to payment inside the app, a messaging app turns into a powerful commerce channel. Both companies are set to enter a full-scale monetization race centered on agentic AI in the second half of 2026.

What It Means for E-Commerce Operators

What the Kakao and Naver experiments show is that the starting point of product discovery is shifting from the search box to messages and conversation. A platform that holds conversation context and relationship data processes everything from recommendation to payment the moment demand arises. If this pattern takes hold, e-commerce operators will increasingly see purchases completed outside their own sites.

The question then becomes experience design that lets a user move smoothly from recommendation to payment. As the report notes, Kakao's challenge is closer to user experience than technology: recommendations must feel natural, the payment process must be short, and users must not sense advertising intervention. To avoid losing this agent-driven demand, operators should consider structuring their product data and connecting to external platforms sooner rather than later.

Conclusion

The Kakao case highlights that the success of AI commerce is starting to be measured not by whether it makes good recommendations but by whether it leads to payment. Special Friend and the place agent are only entry points for finding demand inside KakaoTalk. Only when external partner links and Kanana-based agent commerce produce actual transaction conversion will it be judged to have reached the monetization stage. The agent expected in early August looks set to be the first test of that.