Meta Platforms has been aggressively pushing artificial intelligence into every corner of its product lineup, from chatbots in Messenger to AI-generated image tools on Instagram. But the company’s latest AI-powered shopping assistant, designed to help users find and buy products directly through Meta AI, is drawing sharp criticism from early testers who say the experience is riddled with inaccuracies, dead-end links, and a fundamental misunderstanding of how people actually shop online.
The feature, which allows users to ask Meta AI for product recommendations and receive shoppable suggestions complete with images and prices, represents Meta’s bid to capture a share of the lucrative product-discovery market long dominated by Google and Amazon. Yet according to hands-on testing, the tool appears to be far from ready for the spotlight — and may actually push users further away from Meta’s platforms rather than keeping them engaged.
A Shopping Assistant That Can’t Get the Basics Right
In a detailed review published by Lifehacker, technology writer Pranay Parab described his experience testing Meta AI’s shopping capabilities as deeply frustrating. Parab asked the assistant for a range of product recommendations — from running shoes to tech accessories — and found that the results were consistently unreliable. Product images frequently didn’t match the items described, prices were inaccurate or outdated, and many of the links provided led to pages where the product was unavailable or entirely different from what was shown.
“I won’t be using it again,” Parab wrote bluntly, summarizing an experience that included phantom products, misleading thumbnails, and a general sense that the AI was fabricating shopping results rather than pulling from verified, real-time inventory data. The review highlighted a core tension in Meta’s AI strategy: the company is racing to embed AI features across its apps, but the underlying technology doesn’t yet deliver the reliability that commercial transactions demand.
The Hallucination Problem Comes to Commerce
The issues Parab identified are not unique to Meta. AI hallucination — the tendency of large language models to generate plausible-sounding but factually incorrect information — has been a persistent challenge across the industry. When hallucinations appear in a casual chatbot conversation, the stakes are relatively low. But when they show up in a shopping context, where users are being asked to spend real money based on AI-generated recommendations, the consequences are far more serious.
Meta AI’s shopping feature appears to pull product information from across the web and from Meta’s own Shops infrastructure on Facebook and Instagram, but the synthesis of that data is where things break down. Products that appear to be in stock may not be. Prices displayed by the AI may not reflect current listings. And in some cases, the AI appears to generate product suggestions that don’t correspond to any real listing at all — a particularly dangerous failure mode for a tool designed to facilitate purchases.
Meta’s Broader AI Commerce Ambitions
Meta has been signaling for months that AI-powered shopping is a strategic priority. CEO Mark Zuckerberg has spoken repeatedly about using AI to improve product discovery on Instagram and Facebook, where billions of users already browse content from brands and creators. The company’s Advantage+ shopping campaigns, which use AI to automate ad targeting and creative optimization, have been a bright spot in Meta’s advertising business, generating billions in revenue from e-commerce advertisers.
But there is a meaningful difference between using AI behind the scenes to optimize ad delivery — where Meta has years of data and proven systems — and putting AI in front of users as a direct shopping advisor. The latter requires a level of accuracy and trust that current large language models struggle to provide. Meta’s AI shopping assistant essentially asks users to trust a chatbot with their purchasing decisions, and early indications suggest that trust is being violated almost immediately.
How Competitors Are Handling AI Shopping
Meta is not the only tech giant experimenting with AI-assisted commerce. Google has integrated AI-powered shopping features into its Search Generative Experience, offering product comparisons, price tracking, and visual search capabilities that draw on Google Shopping’s extensive merchant database. Amazon, meanwhile, has introduced Rufus, an AI shopping assistant built into its mobile app that answers product questions and helps users compare items within Amazon’s own inventory.
The key difference is that both Google and Amazon have spent decades building structured product databases with verified pricing, availability, and merchant information. Their AI shopping tools are layered on top of that infrastructure, which provides a factual foundation that constrains the AI’s tendency to hallucinate. Meta, by contrast, has a comparatively thin commerce infrastructure. While Facebook Marketplace and Instagram Shops exist, they lack the depth and rigor of Amazon’s product catalog or Google’s merchant feeds. This means Meta’s AI shopping assistant is working with less reliable source material from the start, making errors more likely.
The Trust Deficit in AI-Powered Recommendations
Consumer trust is the currency of online commerce, and it is extraordinarily difficult to rebuild once lost. Research from multiple firms has shown that shoppers who encounter inaccurate product information — whether wrong prices, misleading images, or broken links — are significantly less likely to return to that platform for future purchases. A single bad experience can undo months of engagement-building.
For Meta, which has already faced years of skepticism from users and regulators over data privacy and content moderation, adding an unreliable shopping assistant to the mix compounds an existing trust problem. Users who try Meta AI’s shopping feature and receive bad recommendations aren’t just disappointed by the AI — they may begin to question the reliability of product-related content across Instagram and Facebook more broadly. This is particularly concerning for the millions of small businesses that depend on Meta’s platforms to reach customers.
Why Rushing AI Features to Market Carries Real Risk
The pressure on Meta to demonstrate AI progress is immense. The company has invested tens of billions of dollars in AI infrastructure, and Wall Street is watching closely for signs that those investments are translating into user-facing products and revenue growth. Zuckerberg has framed 2025 as a year when Meta’s AI efforts will begin to pay off in tangible ways, and the shopping assistant is one of the most visible manifestations of that promise.
But shipping AI features that don’t work well carries its own financial risk. If users learn to distrust Meta AI’s recommendations, they may also become more skeptical of the AI-optimized ads that form the backbone of Meta’s $130 billion-plus annual advertising business. The line between an AI shopping assistant and an AI-targeted ad is thin in the user’s mind, and negative experiences with one can bleed into perceptions of the other. Meta’s advertising partners — the brands and retailers paying for product placement — have a vested interest in ensuring that AI-powered discovery tools are accurate and trustworthy.
What Meta Needs to Fix Before AI Shopping Can Work
For Meta AI’s shopping assistant to become a viable product, several fundamental issues need to be addressed. First, the company needs a more rigorous product data pipeline that verifies pricing, availability, and product details in real time before presenting them to users. This likely means deeper integrations with merchant inventory systems and stricter quality controls on the data that feeds the AI model.
Second, Meta needs to implement better guardrails against hallucination in commercial contexts. When the AI is uncertain about a product’s availability or price, it should say so explicitly rather than presenting fabricated information with false confidence. Transparency about the AI’s limitations would go a long way toward preserving user trust, even if it means the assistant sometimes says, “I’m not sure — here’s a link to check directly.”
Third, the company should consider a more gradual rollout that limits the shopping assistant to categories and merchants where data quality is highest, rather than attempting to cover the full breadth of online commerce from day one. A narrower but more reliable tool would serve users better than a broad but broken one.
The Bigger Picture for AI and Online Retail
Meta’s stumble with AI shopping is instructive for the broader technology industry. The rush to embed generative AI into every product category — from search to customer service to commerce — is producing a wave of features that look impressive in demos but fall apart under real-world use. Shopping, in particular, demands a level of factual precision that current AI models are not consistently able to deliver.
As Lifehacker’s review makes clear, the gap between what Meta AI promises and what it delivers in a shopping context is wide enough to drive users away entirely. For a company betting its future on AI, that gap represents not just a product problem but a strategic vulnerability. The question now is whether Meta can close it before users — and advertisers — lose patience.
WebProNews is an iEntry Publication