The next retail war in South-east Asia: marketplaces vs GPT prompts

[SINGAPORE] E-commerce players in South-east Asia are trying to catch up to Shopee, but consumers might soon do their shopping on artificial intelligence (AI) tools such as ChatGPT instead of online marketplaces.

Known as GPT-based or agentic commerce, this new way of shopping allows consumers to search for product information, dig deeper into reviews, and compare products within their conversations on AI tools.

People can also buy directly on these chatbots. Earlier this week, OpenAI released its Instant Checkout feature in the United States. Initially available for Etsy purchases, the feature will soon be expanded to include Shopify merchants.

Indonesia was the second global market where OpenAI rolled out ChatGPT Go, underscoring the country’s as a testing ground for new products and features – the same playbook that TikTok Shop followed.

As such, it is only a matter of time before “buy it in ChatGPT” becomes available and gains traction in South-east Asia.

So where does this leave brands?

Built for AI discovery

Marketplace fees can consume up to 30 per cent of seller revenue in Thailand and Indonesia when ads and logistics costs are included, said Paul Srivorakul, CEO and co-founder of e-commerce enabler aCommerce.

Given the expense, GPT-based commerce could offer brands some relief if they know how to capitalise on the trend. For example, they could invest in their own websites and craft brand stories with details that AI can crawl and feed to users.

ACommerce estimates that some of the brands it serves get up to 9 per cent of site traffic from ChatGPT and Gemini, with users asking eight to 12 product questions before buying. In some cases, conversion rates could be 4x to 6x higher compared to traditional searches or marketplace clicks.

Large language models (LLMs) are more likely to pull from brands with a strong online presence, points out Toni Ruotanen, head of e-commerce in South-east Asia at marketing firm Havas.

Given how LLMs work, brands’ product descriptions must be discoverable to Google Search and AI platforms. Much like search engine optimisation, answer engine optimisation structures content in a way that allows ChatGPT-like tools to recognise and surface it in response to user queries.

Ruotanen explains that AI tools often gather details from official brand websites instead of e-commerce platforms or other third-party sites. Since answers can vary depending on how a user prompts the AI, brands should have clear and consistent information on their sites so they do not miss out.

“When it comes to product descriptions, it really just comes down to good writing,” said Marvel Gomulya, CEO and co-founder of Covena. His startup helps clients automate sales conversations and transactions through AI assistants on channels such as WhatsApp.

“If humans can read and understand why a product is good, then AI will understand it too,” Gomulya added.

Experts, however, differ on what kind of products should be optimised for AI tools. Some believe AI chatbots are better suited for big-ticket items such as electronics, purchases that require more research and deliberation before a decision is made.

For simple goods, people may rely on affiliates and influencers instead of turning to chatbots. Even so, Srivorakul from aCommerce thinks that fashion and beauty brands as well as fast-moving consumer goods could be a good fit for GPT-commerce since consumers buy them regularly and often seek AI recommendations.

Marketplaces catching up

If users can complete purchases within ChatGPT or other AI chatbots, then people might search less on marketplaces.

Even though that is not possible yet in South-east Asia, Shopee and Lazada have taken steps to ensure their AI can keep up once ChatGPT rolls out the shopping feature here.

Tech in Asia tested AI Lazzie by asking for smart TV recommendations and comparisons. While its answers still appeared generic and lacked detail, the potential is clear, as users could be directed straight to the product pages of the recommended items.

During its 9.9 sale campaign, Lazada said that AI Lazzie helped drive purchases, with orders rising 36 per cent across the region through the chatbot compared to the 6.6 sale. Engagement with AI Lazzie also increased by 53 per cent.

The company is getting ready for agent-to-agent commerce, a Lazada spokesperson told Tech in Asia. This could eventually allow integration with external platforms, making it easier for shoppers to discover its products.

Recently, Google also released an open-source payments standard that lets AI agents send and receive money. This means that once platforms such as Lazada connect their agents externally, those agents could complete payments.

Meanwhile, Shopee said that it is “actively exploring” the use of agentic commerce and conversational AI in shopping. In February, its parent company Sea Group announced a partnership with OpenAI to widen access to AI tools for consumers and small businesses.

Ming Yii Lai, project leader and research manager at Daxue Consulting, notes that AI assistants on e-commerce platforms are already an ongoing trend in China. Both Taobao and JD.com have been developing their AI shopping assistants since 2023, with the latter saying more than 50 million consumers have used its agent.

While Lai acknowledges that GPT-based commerce could threaten e-commerce giants, she thinks they still hold advantages such as established supply chains, payments, and fulfilment capabilities.

In addition, these companies can train their AI tools based on data generated by their users over the years, which general LLMs do not have access to.

As for TikTok Shop, consumers often do not have a strong intention to buy there. This could shield the platform from disruption by GPT commerce, said an executive from Tokopedia.

“People who use ChatGPT already have a certain idea of buying something,” the person said. “No one is going to suddenly ask ChatGPT, ‘What should I buy today?’”

Worth the money?

Unlike Google search that shows sponsored products, some people consider ChatGPT as more of a personal assistant.

If GPT commerce takes off, then brands would have to spend for marketing on AI platforms. But would you trust ChatGPT’s answers if it keeps pointing you towards sponsored products?

Another issue is AI hallucinations. Covena’s Gomulya shares a possible scenario: If AI gives incorrect price information or claims that a smartphone’s camera is excellent when it actually isn’t, this could significantly affect a brand’s ability to retain users.

“If trust from consumers is lost, people will not trust that platform anymore. They won’t go back there to ask for information or product comparisons,” he said.

According to the Lazada spokesperson, AI Lazzie only recommends products within the platform that have verified seller information and customer reviews, rather than from open-ended Internet sources.

While the GPT commerce model is still nascent, the Tokopedia executive expects it to become clearer within a year, especially once OpenAI officially launches its in-platform checkout feature in South-east Asia.

However, the executive does not think “GPT commerce will own most of the market – it’s still a long way off”.

“Even if it does happen, it could still take another three to five years,” he concluded. TECH IN ASIA