Meta has quietly begun testing a shopping feature inside its AI chatbot, moving the company squarely into the contest to turn large language models into commerce engines. The feature, available to a subset of Meta AI browser users in the United States, returns product recommendations in a carousel format with images, brand, price and merchant links, and supplements visual results with short bullet-point summaries.
The bot personalizes suggestions where possible, drawing on Meta's knowledge of a user's location and making gender inferences from a user’s name to tailor recommendations. The test does not include an embedded checkout or payment flow; instead users are directed to merchant sites via links. Mark Zuckerberg has framed this work as part of a wider push to deliver a 'unique personalised experience' across Meta's properties and to build new agent-style shopping tools that surface the most relevant items from businesses in Meta's catalog.
Meta's move amplifies a broader industry shift. Google has already woven its Gemini model into retail partners such as Walmart and Sam's Club and published a Universal Commerce Protocol intended to standardise AI-driven shopping. OpenAI has been expanding ChatGPT's capabilities and commercial footprint, extending premium tiers and piloting ads in the service. Chinese firms are taking alternative but aggressive approaches: Alibaba and Ant Group have deepened collaboration with Google Gemini on cross-border AI commerce, while domestic apps are integrating payment rails and platform services to enable AI-initiated orders for food, travel and retail.
For tech platforms, chatbots are no longer novelty conversational agents but potential portals for discovery, traffic and revenue. Meta's strategy—linking AI recommendations to its social graph and merchant catalog—could let it monetise attention through referral fees, advertising and future native commerce. That promise comes with notable trade-offs: using inferred demographics for personalisation risks user backlash and regulatory scrutiny, and steering commerce via opaque models raises concerns about bias, counterfeit listings and the quality of recommendations.
Market analysts see the current round of product launches as the start of a commercialisation phase for large models. Brokerage commentary and platform metrics point to growing token demand for domestic models and faster iteration on multimodal and agent capabilities. If model providers accelerate integrations with retail ecosystems, advertising, social features and payments, 2026 may be remembered as the year AI features migrated from experiment to mainstream revenue channel for internet platforms.
Looking ahead, expect deeper retailer partnerships, more robust merchant integrations and the gradual addition of checkout and payment functions to these AI shopping tools. Industry standardisation efforts such as Google’s Universal Commerce Protocol will influence interoperability, while regulators will scrutinise privacy practices, data use and competitive impacts. The winners will be those that combine reliable product feeds, trustworthy recommendation mechanisms, smooth payment and fulfilment partnerships—and the ability to do so without alienating users or attracting heavy-handed regulation.