Rakuten Launches AI Shopping Assistant on Its E-Commerce Platform

Rakuten, Japan's largest e-commerce conglomerate, has deployed a custom-built AI shopping assistant across its flagship marketplace, marking one of the most ambitious integrations of generative AI into a major retail platform. The assistant, powered by Rakuten's proprietary large language models, serves the platform's estimated 100+ million registered users in Japan.

The move positions Rakuten alongside Western competitors like Amazon, which launched its own Rufus AI assistant in early 2024, and signals a broader shift toward conversational commerce in the global e-commerce industry.

Key Takeaways

• Rakuten's AI shopping assistant is built on the company's in-house RakutenAI family of language models, not third-party APIs
• The tool provides personalized product recommendations, natural language search, and purchase guidance across Rakuten Ichiba's 56,000+ merchants
• Rakuten invested an estimated $1 billion in AI infrastructure during 2024, including GPU clusters and model training
• The assistant integrates with Rakuten's loyalty ecosystem, including its Super Points program with 100+ million members
• Unlike Amazon's Rufus, Rakuten's system is designed to surface products from independent merchants rather than first-party inventory
• The deployment represents the largest AI shopping assistant rollout in the Japanese market to date

Rakuten Bets Big on Proprietary AI Models

Rakuten's decision to build its AI assistant on proprietary models rather than licensing from OpenAI, Google, or Anthropic reflects a growing trend among major Asian tech companies. The company launched its RakutenAI 2.0 family of models in late 2024, ranging from 7 billion to 70 billion parameters, with specialized fine-tuning for Japanese language understanding and e-commerce applications.

The proprietary approach gives Rakuten full control over model behavior, data privacy, and cost structure. Third-party API costs at scale — potentially millions of dollars monthly for a platform of Rakuten's size — make in-house solutions increasingly attractive for companies with the engineering talent to build them.

Rakuten CEO Hiroshi Mikitani has repeatedly emphasized the company's commitment to AI sovereignty. 'We believe owning our AI stack is as important as owning our platform,' Mikitani stated during a recent earnings call. The company recruited over 200 AI engineers in 2024 alone, many from major Silicon Valley firms.

How the AI Shopping Assistant Works

The assistant operates as a conversational layer on top of Rakuten Ichiba, Japan's largest online marketplace. Users can interact with it through text or voice to find products, compare options, and receive tailored recommendations based on their purchase history and browsing behavior.

Key capabilities include:

• Natural language product search: Users describe what they need in everyday language rather than keyword-based queries (e.g., 'I need a waterproof jacket for hiking in cold weather under $150')
• Cross-category recommendations: The AI suggests complementary products across Rakuten's vast catalog of 300+ million items
• Price tracking and deal alerts: The assistant monitors price fluctuations and notifies users when items drop to their desired price point
• Merchant comparison: Side-by-side evaluation of similar products from different sellers, including shipping times, reviews, and return policies
• Gift recommendations: Personalized suggestions based on recipient profiles and occasion context
• Loyalty optimization: Automatic calculation of the best way to maximize Rakuten Super Points on each purchase

Unlike simpler chatbot implementations, the system maintains multi-turn conversation context, allowing users to refine their searches iteratively. A shopper can start by asking about running shoes, narrow down to trail running shoes for wide feet, and then ask the assistant to compare the top 3 options — all within a single conversation thread.

Rakuten's AI Strategy Differs From Amazon's Approach

The comparison to Amazon's Rufus assistant is inevitable, but the two systems differ in fundamental ways. Amazon's Rufus, launched across the U.S. marketplace in 2024, primarily draws on Amazon's vast first-party product data and customer reviews to answer shopping questions.

Rakuten's assistant, by contrast, operates in a marketplace-first model where the platform hosts independent merchants rather than competing with them through first-party sales. This creates a different optimization challenge: the AI must balance user satisfaction with merchant fairness, ensuring that recommendations don't systematically favor larger sellers over smaller ones.

Rakuten has implemented what it calls a 'democratic recommendation engine' that factors in merchant diversity alongside relevance scores. Small and medium-sized sellers receive algorithmic support to ensure visibility, a feature that Rakuten says distinguishes its approach from competitors.

This merchant-friendly approach could prove strategically important. Rakuten Ichiba hosts approximately 56,000 merchants, many of them small businesses that depend on the platform for their livelihood. An AI system that funnels all traffic to a handful of large sellers would undermine the marketplace ecosystem.

The $1 Billion AI Infrastructure Investment

Rakuten's AI deployment rests on a substantial infrastructure investment. The company has reportedly spent approximately $1 billion on AI-related infrastructure since 2023, including dedicated GPU clusters, data center expansions, and partnerships with chip manufacturers.

The company secured a significant allocation of NVIDIA H100 GPUs and has also explored custom silicon solutions to reduce long-term compute costs. Rakuten's cloud division, Rakuten Cloud, now operates AI-optimized data centers in Tokyo and Osaka, with plans to expand to Singapore and India.

Training the RakutenAI models required processing massive datasets of Japanese-language product descriptions, customer reviews, and transaction histories. The Japanese language presents unique challenges for AI models, including three writing systems (hiragana, katakana, and kanji) and complex honorific structures that affect product descriptions and customer service interactions.

Rakuten claims its models outperform general-purpose alternatives like GPT-4 and Claude on Japanese e-commerce tasks by 15-20% on internal benchmarks, though independent verification of these claims remains limited.

Industry Context: AI Reshapes Global E-Commerce

Rakuten's deployment fits into a broader transformation of the global e-commerce landscape. Major platforms worldwide are racing to integrate generative AI into the shopping experience.

Shopify launched its AI assistant Sidekick for merchants in 2024, while eBay has deployed AI-powered listing tools and image search capabilities. In China, Alibaba's Taobao and JD.com have both rolled out AI shopping assistants powered by their respective in-house models.

Analysts at McKinsey estimate that AI-powered personalization in e-commerce could unlock $400 billion to $660 billion in annual value globally by 2027. Conversion rate improvements of 10-30% are commonly reported by platforms that implement AI-driven product discovery.

The trend reflects a fundamental shift in how consumers interact with online marketplaces. Traditional keyword search and category browsing are giving way to conversational commerce, where AI mediates the entire shopping journey from discovery to purchase.

For Rakuten, the stakes are particularly high. The company faces intensifying competition in Japan from Amazon Japan, Yahoo Shopping (owned by Z Holdings), and emerging social commerce platforms. AI-powered shopping experiences could become a key differentiator in this crowded market.

What This Means for Businesses and Developers

Rakuten's deployment carries significant implications for multiple stakeholders in the e-commerce ecosystem.

For merchants on the platform, the AI assistant creates both opportunities and challenges. Sellers who optimize their product listings for AI interpretation — with detailed descriptions, structured data, and comprehensive specifications — will likely see improved visibility. Those relying on keyword-stuffing or SEO tricks may find their tactics less effective against an AI that understands natural language intent.

For competing platforms, Rakuten's move raises the bar for what consumers expect from online shopping. Smaller e-commerce operators may need to integrate third-party AI solutions from providers like Algolia, Constructor, or Google Cloud Retail AI to remain competitive.

For AI developers, Rakuten's proprietary model approach validates the business case for domain-specific language models. While general-purpose models from OpenAI and Anthropic dominate headlines, purpose-built models trained on specialized datasets can deliver superior performance in specific verticals.

The deployment also highlights the growing importance of multilingual AI capabilities. As AI shopping assistants expand globally, models that excel in languages beyond English will command premium value.

Looking Ahead: Rakuten's AI Roadmap

Rakuten has signaled that the shopping assistant is just the beginning of a broader AI integration across its ecosystem. The company operates in financial services, mobile telecommunications, travel, and digital content — all areas ripe for AI enhancement.

Upcoming developments reportedly include:

• Visual search integration: Users will be able to upload photos to find matching or similar products
• AR try-on features: AI-powered augmented reality for clothing, accessories, and cosmetics
• Predictive inventory alerts: The assistant will anticipate user needs based on purchase patterns and proactively suggest reorders
• Cross-platform integration: Extending the AI assistant to Rakuten Travel, Rakuten Books, and Rakuten Mobile
• International expansion: Plans to bring the assistant to Rakuten's operations in select international markets by late 2025

The company is also exploring agentic AI capabilities, where the assistant could autonomously execute multi-step tasks like comparing prices across merchants, applying coupons, and completing purchases — all with a single user instruction.

Rakuten's AI shopping assistant represents a significant milestone in the convergence of large language models and e-commerce. As the technology matures and user adoption grows, it could fundamentally reshape how hundreds of millions of consumers discover and purchase products online. The race to build the definitive AI-powered shopping experience is accelerating, and Rakuten has made clear it intends to compete at the highest level.