From Washington: In a development with implications stretching well beyond Silicon Valley, Meta has begun testing an artificial intelligence shopping assistant inside its Meta AI platform, pitching itself into one of the most commercially loaded contests in the tech industry today.
The feature, which Bloomberg first reported on Tuesday, is currently visible only to a select group of US-based users accessing Meta AI through a desktop browser. Those with access see a "Shopping research" button inside the query text box. When they enter a product-related request, the assistant generates a carousel of product cards showing images, prices, brand information, and links to the relevant e-commerce site, according to Engadget. The chatbot also provides a brief explanation for each recommendation.
Where it can access profile data, including a user's gender and location, Bloomberg reports the tool tailors results accordingly, surfacing, for example, women's puffer jackets from stores that ship to New York based on a tester's profile. Users cannot complete a purchase inside Meta AI itself; they are directed to external merchant sites to check out. A Meta spokesperson confirmed the test is underway but declined to provide a timeline for any broader rollout.
The move is hardly unexpected. Mark Zuckerberg told investors during a January earnings call that Meta's new agentic shopping tools would help people "find just the right, very specific set of products" from businesses in Meta's catalogue. The speed of execution is striking, though: industry observers noted that just 33 days passed between that earnings call and the appearance of a live, testable feature.
Meta's competitors have not been idle. OpenAI launched a dedicated shopping research feature for ChatGPT in late November 2025, timed to coincide with Black Friday. Google followed with its own shopping tools for Gemini around the same time, and Perplexity released a comparable assistant concurrently. ChatGPT and Gemini have since gone further, building retailer integrations with major chains, while OpenAI is developing an in-chat checkout capability through its Instant Checkout programme. Meta has not confirmed whether it plans to enable in-chat purchases or whether it collects commissions on merchant referral clicks.
The commercial stakes are considerable. The global AI shopping assistant market is projected to grow from US$4.26 billion in 2025 to US$36.38 billion by 2034, a compound annual growth rate of around 26.8 per cent, according to industry analysis cited by AInvest. McKinsey forecasts that agentic commerce could drive between US$900 billion and US$1 trillion in US retail revenue alone by 2030. For Australian exporters and retailers with a US presence, the question of which AI platform becomes the default product discovery layer carries real commercial weight.
There are legitimate questions here that go beyond commercial competition. Meta's proposition rests on the richness of its user data, spanning behavioural signals, social connections, and demographic information across Facebook, Instagram, and WhatsApp. That depth gives it a potential personalisation advantage over rivals, but it also concentrates enormous influence over product discovery inside a single corporate ecosystem. Privacy advocates and consumer groups in Australia and elsewhere have long raised concerns about how Meta handles personal data; an AI that actively leverages that data to steer purchasing decisions will almost certainly intensify that scrutiny.
Proponents of the technology argue the benefits are real. Consumers face genuinely overwhelming choice online, and a well-designed AI assistant that surfaces relevant products quickly, explains its reasoning, and links directly to reputable merchants offers tangible convenience. PYMNTS Intelligence data suggests that more than 60 per cent of Americans used AI for some purpose in the past year, and that as of January 2026, around 41 per cent of consumers had used dedicated AI platforms specifically for product discovery.
The honest assessment is that AI-powered shopping sits at the intersection of competing legitimate interests. The efficiency gains for consumers are real. The concentration risks, for retailers who depend on algorithmic visibility, for regulators overseeing market competition, and for users whose data is doing the commercial heavy lifting, are equally real. Meta has not confirmed how it plans to monetise the feature, a question that matters enormously for how the tool is designed and whose interests it ultimately serves. Reasonable scepticism about a product still in limited testing is warranted, but so is recognition that this technology is arriving regardless of who builds it first. The more pressing question for policymakers, including Australian regulators who oversee how global platforms operate domestically, is what transparency and accountability standards should apply before these tools reach mass deployment.