Consumers increasingly rely on AI shopping assistants to make sense of crowded marketplaces. But a new study suggests that when those assistants spot something misleading, they may not always share that information with the person asking. Researchers at Columbia Law School’s Center for Law and the Economy found that Amazon’s Alexa for Shopping and Walmart’s Sparky chatbot are both capable of detecting fake 'Made in USA' labels. Yet the same tools reportedly remain silent about the problem, and in some cases appear to avoid the topic entirely.
What the study found
The Columbia researchers tested both AI assistants directly with a series of product-specific questions. According to the results, each assistant was able to identify mismatches between a bold 'Made in USA' claim in the product listing and a contradictory country-of-origin detail buried elsewhere on the same page. That matters because a listing can technically contain both statements, but a shopper scanning the page quickly may only register the patriotic cue.
The study also found that this kind of fraud is fairly common across both platforms. The researchers encountered numerous examples where the main title or prominent image emphasized American origin while the product details or shipping information pointed to a different country. In those situations, the AI assistants recognized the conflict. However, they did not volunteer that information unless researchers asked a very specific follow-up question, and even then the responses were incomplete.
How the experiment worked
The researchers used a methodology designed to mimic a typical consumer interaction. They asked each AI assistant about a set of products that had both a prominent 'Made in USA' label and a less visible country-of-origin note pointing to another country. In some cases, the conflicting information appeared in the same product title. In others, the contradiction was hidden in a specifications table or a shipping policy section. The AI assistants were able to parse those details and identify the discrepancy when directly asked about it.
That close reading is exactly the kind of task that should make AI useful for shoppers. A human browsing a product page might not scroll through every field to verify whether the 'Made in USA' claim appears elsewhere with a caveat. An AI assistant, by contrast, can scan the entire listing in seconds. The fact that the technology works only when prompted, and even then is not fully forthcoming, points to a deliberate design choice rather than a technical shortcoming.
How the assistants responded
When researchers asked the bots why their companies were not acting on the issue, neither pointed to any technical limitation. Amazon’s Alexa reportedly admitted that the harm to American brands is real, but explained that the issue stays easier to ignore until it costs Amazon something directly. That response suggests a business calculation rather than an inability to address the problem.
Walmart’s Sparky was even more blunt. The chatbot stated outright that its silence reflects a business calculation rather than a legal one. In other words, the Walmart assistant seemed to acknowledge that false 'Made in USA' labels were detectable but not worth changing the shopping experience over. The researchers also noticed that Amazon’s assistant readily answers questions about products made in China, while blocking similar questions about 'Made in USA' claims entirely. That asymmetry is striking because it shows the technology can engage with country-of-origin topics when the platform chooses to allow it.
The business of silence
Neither company denied that the AI could detect the false claims. Instead, the researchers said both platforms appeared to make a calculated choice. Amazon’s assistant suggested that the problem was not urgent until it caused direct harm to Amazon. Walmart’s assistant was similarly explicit about the commercial logic. This kind of response is particularly troubling because it suggests that AI tools are being tuned to protect the platform's interests rather than the interests of consumers.
The platform economy has long relied on third-party sellers to fill catalogs with enormous variety. That model brings benefits, but it also creates challenges for quality control. Retailers have built systems to police counterfeits
Source: Digital Trends News