AI, E-commerce, and the Battle for Discovery: A China-Europe Comparison

Introduction

For decades, keywords dominated internet rankings as a crude proxy for user intent. Classic algorithms that preceded Artificial Intelligence (AI), such as Google’s PageRank, matched user queries to website keywords. They then displayed the most authoritative webpages (those most cited by prominent domains) and those most relevant by language and location. These algorithms function mechanically, matching keywords without semantic understanding.

As technology progressed, new algorithms emerged. Word embeddings are the backbone of any Large Language Model (LLM), which today power the most famous AI tools. These technologies understand a word’s context within text and its relative distance from other words, allowing them to distinguish, for instance, between ‘bank’ in ‘river bank’ and ‘commercial bank’. While initially used to improve search engine algorithms, these models evolved into today’s conversational AIs, such as ChatGPT, DeepSeek, Qwen, Gemini, or Claude. Eventually, these chatbots became agentic, in other words capable of using tools such as web search. The ability of these models to understand vague requests and anticipate user needs is making internet browsing easier and is shifting consumer behaviour globally. Today, increasing numbers of users rely on generative AI for product searches.

Major technology companies responded swiftly to these behavioural shifts. Google first introduced the ‘Search Generative Experience’ (SGE), which later evolved into AI Overviews: a Gemini-powered feature that combines information from multiple sources into a single summary answer. E-commerce platforms that function as product search engines experienced similar transformations, with AI improving personalisation and the ease of quickly finding the right product. The web is transitioning from ranking logic – where visibility depends on ranking to a top position in the results page – to recommendation logic, where AI selects a curated few. Data shows that only 17% of SGE-enabled queries reference pages currently rank first to third. 

This evolution is progressing at different rates across the world. In China, AI-driven product discovery seems embedded in all-in-one platforms, such as Taobao. In Europe, discovery still relies on external search engines such as Google. This contrast reveals two distinct paths toward the future of digital commerce.

AI in Chinese E-commerce 

Chinese e-commerce companies are expanding significantly across the EU. Companies such as Temu and Shein have already penetrated EU shopping habits. Others, such as AliExpress and Miravia for business-to-consumer (B2C) and Alibaba.com for business-to-business (B2B), are expanding steadily in Southern Europe. Meanwhile, Jingdong (JD) is investing in logistics infrastructure to enable future market entry. Chinese e-commerce platforms have launched AI shop assistants enabling users to find products by describing tasks (for example, ‘I need to fix a water leak.’) or contexts (such as ‘Gift for my girlfriend’s birthday, budget 1,000 CNY’). In China, these companies operate through ‘super-apps’, vertically integrating the value chain and controlling the user's consumption experience from discovery to final delivery. Because ‘super apps’ are so vertically integrated and follow clients from the very beginning of the shopping journey to the very end, it is easier for Chinese platforms to develop AI systems that learn users’ tastes and experiences in depth.

A prominent example is Alibaba’s Accio. Launched in late 2024, it was branded as the ‘world’s first AI agent for global trade’. Upon receiving requests such as ‘I want to build an indoor ski park, what do I need?’, Accio conducts comprehensive internet searches to identify required products, suggests Alibaba.com suppliers, and displays comparison tables. Data shows that in China, Accio increased product conversion rates by 20-30%. Accio’s product-discovery method surpasses keyword matching by analysing entire product pages, including photos and videos, to better align with user expectations. This appears in the Product Information Score (PIS) – a product page evaluation metric – which increasingly rewards content richness (images, specifications), clarity (shipping information, pricing), and trendiness over keyword stuffing and basic categorisation.

The Alibaba Group introduced similar features for its B2C operations. For instance, Taobao launched ‘AI万能搜’ (AI Universal Search) which is similar to Accio, as well as its so-called ‘AI Assistant’, which facilitates the buyers’ decision-making through conversation. These tools retain previous client discussions to tailor recommendations, request clarifications, and may eventually handle purchasing and after-sales service. This transformation reshapes e-commerce exposure dynamics, enabling niche products to emerge whilst restructuring advertising and product listing protocols.

AI in European E-commerce

In the EU, e-commerce is dominated by U.S. firms, and there is no equivalent to ‘super-apps’ in the Chinese sense. European customers primarily search for goods on Google, Amazon, and similar platforms.

In the U.S. and the EU, Amazon was, until recently, the primary platform for finding and comparing products. In 2024, over 50% of U.S. customers reported starting their online shopping searches there. The e-commerce giant has a tradition of using machine learning and AI for relevance. Currently, the company prioritises personalisation over conversational search agents. Amazon uses AI to modify product titles and descriptions, tailoring them to individual user queries and behaviour based on the product details entered by the seller. If a customer, for instance, frequently searches for gluten-free foods and then looks for cereal, Amazon’s AI can dynamically insert the term ‘gluten-free’ into the cereal product titles, even if the seller’s original title had ‘gluten-free’ buried in the product features. Behind the scenes, Amazon employs a two-LLM system: one model rephrases product content to emphasise likely user priorities, while a second evaluates output accuracy. This process helps avoid misleading the user by highlighting features the product lacks. Given Amazon’s Alexa platform and experimentation with its chatbot Rufus, the company may eventually introduce conversational AI assistants that look similar to those seen on the Chinese competitors.

Europe hosts other prominent e-commerce players such as Zalando, eBay, and Vinted, which progressively integrated AI to enhance relevance and product recommendations. Zalando recently deployed an AI assistant to give users fashion advice and understand user intents; eBay followed a similar approach. However, these platforms have a narrower market scope (fashion, vintage) and data on their AI performance remains unavailable. While perhaps less advanced than China’s in-platform AI assistants, product discovery in the EU is shifting from keyword-based ranking towards AI-curated suggestions, drawing closer to Chinese development patterns.

Google instead uses its ‘AI Overview’ to generate product comparisons within search results, displaying products directly so that users need not click external websites. In practice, Google is moving down the marketing funnel towards product-discovery and comparison functions traditionally occupied by marketplaces. In the future, platforms such as ‘AI Overview’ by Google or ChatGPT, which operate outside conventional e-commerce platforms, may enable users to complete purchases directly from AI-generated product recommendations as their agentic capabilities mature. Recent developments already point in this direction: Shopify, a leading platform in Europe for building independent brand-owned e-commerce sites, has recently launched the ‘Universal Commerce Protocol’ (UCP), which allows products to be purchased directly by AI agents. While this offers brands a viable alternative to selling through established marketplaces such as Amazon or eBay, it also raises the stakes for optimising brand websites: to be surfaced by AI systems, these sites must be structured and presented in ways that AI agents can easily interpret, evaluate, and recommend.

Implications for European Businesses

E-commerce marketplaces remain excellent entry points for European exporters, reducing the need to find local distributors, build proprietary platforms and AI tools, or invest heavily in digital marketing. The recent expansion of Chinese platforms in Europe further diversifies the options for reaching international customers. 

AI is reshaping product discovery on two fronts simultaneously, namely within e-commerce platforms and through external AI-driven search. The Chinese model demonstrates how integrated AI assistants reshape in-platform behaviour: tools such as Accio and Taobao’s AI search interpret user goals and generate complete solution sets rather than matching keywords. In the meantime, Europe is witnessing a parallel transformation of external search, with Google’s AI Overviews and general-purpose chatbots answering questions whilst bypassing traditional rankings. The former represents ‘platform-native AI’, the latter ‘search-based AI’. Future e-commerce success may depend on adapting to both paradigms. Search-based AI (Google’s AI Overview and ChatGPT, among others) requires webpage optimisation beyond traditional Search Engine Optimisation (SEO). AI-friendly websites must contain useful content that addresses user questions. AI prefers structured content such as tables and lists from which data can be easily extracted, with particular attention to paragraphs that make sense independently and sound conversational in tone. Additionally, websites should maintain frequently asked questions sections (structured using JSON-LD) that address topic-specific sub-questions, as AI agents generate additional queries beyond user inputs. Lightweight navigation also matters, since AI prioritises energy efficiency. Further, using HTML rather than JavaScript for page content facilitates web scraping for AI systems. Finally, sustained content marketing – defined as publishing original, domain-specific articles and insights on owned and social channels – reinforces perceived expertise. This improves a company’s visibility and credibility both for AI-driven discovery systems and for human readers.

Platform-native AI optimisation depends on marketplace-specific logic and enables AI to navigate product pages without loading or scraping constraints. Beyond complete product information, companies should emphasise images and videos, given AI’s expanding multimodal processing capabilities. Sellers should also explicitly describe concrete product use cases – that is, the situations or needs a buyer may have when searching for a product. Providing this contextual information helps AI-driven recommendation systems to better match products to user intent, increasing the likelihood that the item will be recommended.

Finally, advertising campaign effectiveness is undergoing a significant transformation. AI will inevitably reduce both click-through rates and the importance of Search Engine Ads (SEA), and this effect varies across query types. However, this trend is gradual: currently, traditional browser-based search remains dominant. Despite that, new forms of AI-related advertising are emerging, and businesses will have to consider how to rebalance their digital marketing investments based on the role AI-powered search plays for their target audience.

In conclusion, comparing Western and Chinese AI-based search approaches provides a more comprehensive account of the ongoing digital commerce transformation. In China, product discovery is primarily conducted within e-commerce platforms’ own applications, favouring AI systems optimised for in-app search and recommendation. By contrast, the European ecosystem is gradually moving towards a hybrid model that blends on-platform search with traditional search engines, now increasingly mediated through conversational, chatbot-style interfaces. Companies that render their content and product listings easily interpretable and relevant for AI systems will benefit most from the evolution of product discovery. 

The views expressed in this article belong to the author(s) alone and do not necessarily reflect those of European Guanxi.

ABOUT THE AUTHORS

Giacomo Savarese (司马睿) is the first author of this article. He is from Tuscany and earned his Bachelor’s and Master’s degrees from Bocconi University in Milan, Italy where he specialised in AI and marketing. He has experience with Chinese tech firms and worked for companies such as Alibaba.com. He is also passionate about modern history and geopolitics, and wrote his bachelor’s thesis on the Chinese strategy in Central Asia. He currently works at Accenture Song as an AI specialist.

Jiaming Ma (马嘉明) was born in Beijing, China. She is a biology and economics double major graduate from Dartmouth College in the U.S. She is currently pursuing an MD degree at the Yale School of Medicine. She has a strong interest in emerging technologies. She wrote her undergraduate thesis on CRISPR gene editing and phylogeny analysis. In medical school, she is currently engaging in research related to medical AI applications.

This article was edited by Robin Millet and Isabell Raue.

Featured Image:  UnionPay Mobile Payment with Online Shopping / Julio Lopez  / Free for use  / Pexels