TL;DR
- New Feature: Meta is quietly testing an AI-powered shopping research tool inside Meta AI that surfaces product carousels for U.S. browser users.
- How It Works: The feature runs as a dedicated shopping mode showing product images, prices, and recommendations, though the checkout button remains non-functional.
- Competitive Context: Meta enters the AI shopping race months after ChatGPT launched in November 2025 and Google introduced its Universal Commerce Protocol in January 2026.
- Strategic Advantage: Meta brings Facebook Shops, Instagram Shopping, and behavioral data from 3.2 billion daily active users as unique advantages over AI-only rivals.
When a New York shopper asked Meta AI about puffer jackets this week, the chatbot returned a personalized carousel of results, the first public glimpse of a shopping research feature Meta is quietly testing to challenge ChatGPT and Gemini. Bloomberg first reported the test on Tuesday. The rollout covers a small group of U.S. users, and the feature has not been announced publicly by Meta.
Meanwhile, Meta is the last of the three major AI chatbot platforms to enter this space, arriving with distribution advantages via Facebook, Instagram, and WhatsApp that neither rival can match natively. As AI assistants become the default starting point for more online activity, the platform that intercepts shopping intent earliest gains meaningful leverage over how consumers reach purchase decisions.
How Meta AI’s Shopping Feature Works
That early glimpse is built on a deliberate structural choice. When a U.S. user enters a shopping prompt into Meta AI through a browser, the assistant initiates a “thinking phase” before surfacing a product carousel of results.
Those cards display product images, brand information, prices, and bullet-point recommendations tailored to the query. Users who trigger the dedicated shopping mode – which appears as a distinct option in the browser interface rather than a standard response type – receive a structured shopping experience separate from general chat.
Clicking a product card opens a side panel with detailed descriptions, additional visuals, and quick-purchase options. The buy button and checkout flow remain non-functional in the current build; completed purchases route to external retailer websites.
Meta has not disclosed whether it has commission arrangements with those retailers or how advertisers are prioritized in the carousel alongside organic recommendations. The deliberate separation of a dedicated shopping mode from general chat – rather than treating shopping as just another response type – indicates Meta is building a commerce destination, not a conversational add-on.
The inactive buy button and undisclosed commission structure point to a company still resolving its monetization approach before wider rollout. That pattern is consistent with how Meta has historically tested ad products at small scale before embedding them platform-wide.
Currently browser-only and restricted to U.S. users, the feature was identified independently by TestingCatalog before Bloomberg’s report. Shopping appears as a dedicated mode rather than a response type activated by prompt content alone – meaning users must explicitly trigger it rather than receiving shopping results inside standard chat responses. Mobile users do not currently have access.
Entering the AI Shopping Race
Despite that measured rollout, the competitive field Meta is entering has already taken shape. Meta enters as the third major AI platform to integrate shopping research into a general-purpose chatbot. OpenAI’s shopping feature in ChatGPT launched in November 2025, built on GPT-5 Mini and focused on product comparison rather than transaction completion.
OpenAI paused instant checkout at launch, prioritizing research accuracy over purchase facilitation. Google’s Universal Commerce Protocol followed in January 2026, enabling Gemini to compare products, check prices, and complete purchases via Google Pay through a universal checkout standard.
Moreover, the field has continued to expand beyond those two platforms. ChatGPT and Gemini are both deepening retailer integrations, with Walmart and Sam’s Club among confirmed partners. OpenAI is developing in-ChatGPT purchases through PayPal, while Microsoft is testing a Copilot Checkout feature and Amazon’s Alexa+ supports conversational shopping with voice-initiated product selection.
Google had previously launched an agentic checkout that completes purchases for shoppers through its AI shopping agents, part of a broader agentic commerce push. The rapid proliferation of AI shopping integrations signals growing industry consensus that conversational product research will reshape how consumers enter the purchase funnel. Traditional search engines face displacement as the default starting point for retail spending decisions.
Prior Coverage and Competitive Timeline
The four-month gap between OpenAI’s November 2025 launch and Meta’s current test carries concrete consequences beyond brand perception. Retailer partnerships, data-sharing agreements, and checkout integrations tend to consolidate around early movers – Walmart and Sam’s Club have already committed to ChatGPT and Gemini integrations. Each additional month narrows the pool of premium retail partners available to Meta on equivalent terms.
In contrast, Meta arrives later but with pre-existing commerce infrastructure – Facebook Shops, Instagram Shopping, and years of behavioral data from billions of users reflecting purchase intent at scale – that purpose-built AI shopping tools cannot replicate quickly. That compounding disadvantage is one distribution scale alone may not offset, making the timing of Meta’s entry a strategic liability as much as an opportunity.
Meta AI’s Commerce Strategy
That infrastructure underpins ambitions that extend well beyond the current test. According to Meta’s most recent earnings disclosures, the company’s family of apps reaches 3.2 billion daily active users. The shopping research test sits within a broader push to transform Meta AI from a conversational assistant into a platform that captures commercial intent at the moment it forms.
On a recent investor call, Zuckerberg signaled a deliberate strategic push, telling investors that upcoming AI models carry implications for commerce and agentic shopping. TestingCatalog reported those remarks and provided additional context on the Meta AI shopping research feature:
“Zuckerberg recently told investors that the company’s upcoming AI models and products will have ‘implications for commerce,’ including agentic shopping tools designed to help consumers discover products from Meta’s business catalog.”
TestingCatalog (via TestingCatalog)
Furthermore, TestingCatalog found that Meta AI routes some search queries through Google’s Gemini 3 internally, suggesting the final product may run on a multi-model system rather than Meta’s own Llama exclusively. Meta also acquired Manus, an autonomous AI agent technology company, according to reports, to accelerate its conversational platform capabilities and social commerce ambitions.
The reported Gemini 3 routing – combined with the Manus acquisition – indicates Meta is assembling a layered AI stack rather than racing to ship a single vertically integrated product. For the shopping feature specifically, multi-model routing could enable Meta to use specialized models for product ranking while reserving its own infrastructure for personalization against its behavioral data.
What Analysts Are Watching
Beyond the product mechanics, Wall Street is already assigning value to Meta’s commerce play. This type of agentic commerce could shift how consumers discover products online, giving Meta more influence over the browsing-to-purchase journey.
Wells Fargo analyst Ken Gawrelski noted that “compute capacity – the total computer processing power available – is a determining factor of success” for AI-driven commerce. He raised his Meta price target to $856 from $844 on February 23, 2026, maintaining an Overweight rating with shares trading at $648.18.
Analysts project hyperscale data center power consumption will reach 98 gigawatts by 2027, reflecting $860 billion in capital expenditures across the sector – the infrastructure backdrop against which Meta’s agentic commerce ambitions will be tested.
Gawrelski’s $12 price target increase – issued before any shopping revenue has materialized – reflects analyst confidence that Meta’s behavioral data and merchant infrastructure represent monetizable assets the market has not yet fully priced. Meta’s reported Q2 2025 performance gains, including a 5% improvement on Instagram Reels from AI models like GEM, demonstrate that its infrastructure investments already produce measurable business results. That track record lends credibility to its agentic commerce ambitions.
For the hundreds of millions of consumers who use Meta’s apps daily, the stakes are concrete. Shopping decisions that once began on Google Search or a retailer site may increasingly originate inside a chatbot embedded in their social feed.
Whether those recommendations reflect genuine search quality or advertiser prioritization – a question Meta has yet to answer publicly – will determine whether users adopt the feature or treat it as another ad surface. Meta has not announced a timeline for expanding beyond U.S. browser users or when mobile access will be enabled.