How Amazon and Walmart’s AI assistants are changing product discovery
Subscribe to our free newsletter today to keep up to date with the latest retail news.
Online shopping is moving beyond a model built primarily around search boxes, filters and pages of product results, as consumers increasingly use AI assistants to describe what they want and receive a narrower selection of products based on those requirements.
Amazon and Walmart are developing their shopping experiences around this shift, with both retailers introducing AI assistants designed to help consumers research, compare and select products. Amazon’s Alexa for Shopping, previously called Rufus, uses generative and agentic AI to support product discovery and evaluation, including personalized recommendations based on conversational context and shopping activity. The assistant can perform transactional tasks such as adding products to a cart and automatically purchasing an item when it reaches a specified price.
Walmart is taking a comparable approach with Sparky, its generative AI shopping assistant that can compare products, synthesize reviews and recommend purchases based on customer requirements. Walmart has said the technology is being developed to handle a wider range of activities, including product reordering and service booking.
For manufacturers, these capabilities introduce another intermediary between their products and prospective customers because visibility increasingly depends on whether an AI system can interpret product information, determine its relevance to a request and include it among its recommendations. Country of origin provides a useful example of the difficulties that can arise when an important purchasing criterion is not consistently available or interpreted across ecommerce systems.
Made in USA highlights gaps in product data
Research published in July by Columbia University’s Center for Law and the Economy examined how Amazon’s Alexa for Shopping and Walmart’s Sparky responded to questions about country of origin. The researchers reported that both systems could detect potentially questionable Made in USA claims, but they also identified differences in how the assistants responded when consumers attempted to find American-made products.
According to the center, Amazon’s assistant declined or blocked some questions about Made in USA products while answering comparable questions concerning goods made in China. Researchers also reported that changing the wording of a query could affect whether the assistant returned country-of-origin information, suggesting that access to the information was not always consistent across different prompts.
The Wall Street Journal reproduced some of the reported behavior and found that Alexa for Shopping said it lacked access to information needed to identify certain products made in the USA, while a modified prompt subsequently produced products identified as US-made. Walmart’s Sparky initially declined one request to assess Made in USA claims before responding to a similar request, while Amazon disputed the report’s characterization and said it displays country-of-origin information when that information is available.
The findings do not establish that retail AI systems are deliberately suppressing American-made products, since such a claim would require evidence about platform policies, model behavior and system design that outside researchers may not possess. They do identify a practical problem for ecommerce: AI shopping assistants can produce inconsistent responses when consumers ask about a product characteristic that may influence their purchasing decisions.
Country of origin is particularly complex because Made in USA is not simply a descriptive marketing phrase. Under the Federal Trade Commission’s Made in USA Labeling Rule, marketers making an unqualified Made in USA claim must be able to substantiate that a product is “all or virtually all” made in the United States.
The FTC’s enforcement activity shows that these requirements have practical consequences for manufacturers and sellers. In July, the agency sent warning letters to seven companies concerning potentially misleading Made in USA claims and another company concerning potentially misleading Made in Texas claims, with products cited by the agency including industrial laser machinery, coordinate measuring machines and drums.
An AI shopping system cannot reliably treat every reference to “American,” “USA” or “US company” as evidence that a product meets the federal standard, yet withholding country-of-origin information can also reduce the usefulness of the assistant when a marketplace possesses information relevant to the customer’s request. Retailers therefore face the technical and regulatory challenge of determining which origin information can be trusted and how it should be presented through AI interfaces.
Manufacturers may need to optimize product data for AI
Traditional ecommerce encouraged manufacturers to develop a familiar set of practices around digital visibility, including searchable product titles, relevant keywords, structured marketplace listings, customer reviews and conversion optimization. These practices generally assumed that consumers would receive a set of search results and then evaluate individual products themselves.
Agentic commerce changes part of that process because an AI assistant can interpret a request, evaluate available information, compare products and present a considerably smaller selection. A consumer asking for durable work boots made in the USA for less than $200, for example, may rely on the assistant to identify which products meet all three criteria rather than independently reviewing dozens of listings.
As more product evaluation is delegated to AI systems, the information available to those systems becomes more important to manufacturers. Research into agentic ecommerce has shown that AI shopping systems can respond differently to factors such as product position, price, ratings, reviews, sponsored labels and platform endorsements, which means product visibility may depend partly on how individual systems interpret and rank marketplace information.
A 2025 study using a simulated marketplace examined AI shopping agents under changes in product position, pricing, ratings, reviews, sponsored labels and platform endorsements. The researchers identified differing position effects among models and found that AI agents tended to penalize sponsored labels while responding positively to platform endorsements, while changes to product descriptions could affect market share when sellers optimized listings for AI buyers.
Research published in July 2026 also suggests that the mechanisms used by agents to query and rank products can affect exposure and purchasing outcomes. For manufacturers, these findings raise questions about whether existing ecommerce optimization practices will be sufficient when product selection increasingly involves machine interpretation rather than direct comparison by consumers.
Product information may need to become more structured and machine-readable, particularly for attributes such as manufacturing location, assembly location, component origin, certifications, materials and warranty conditions. Manufacturers that compete on these characteristics may have difficulty communicating their differentiation if the relevant information is buried in descriptive copy, recorded inconsistently across marketplace fields or unavailable to the AI system making the recommendation.
This issue could be particularly relevant for smaller US manufacturers whose products carry a price premium linked to domestic production. If a US-made product costs $170 while an imported alternative costs $110, an AI assistant that places significant weight on price may favor the lower-cost product unless the consumer specifies domestic manufacturing as a priority and the system has reliable origin data with which to evaluate that requirement.
The quality and structure of product data could therefore become a more significant component of ecommerce competitiveness, requiring manufacturers to consider not only how information appears to customers but also how accurately it can be interpreted by automated shopping systems.
AI shopping increases the value of verified product information
The questions surrounding Made in USA claims reflect a wider issue that extends beyond country of origin because consumers may ask AI assistants about manufacturing practices, repairability, materials, environmental characteristics, sourcing standards and other attributes that are more difficult to verify than price, availability or customer ratings.
A marketplace can provide an AI assistant with a product’s current price through a structured data field, whereas determining whether the same product satisfies a legally meaningful country-of-origin standard requires a more detailed set of information. Manufacturers may need to provide that information in standardized formats, marketplaces may need systems for organizing and validating it, and regulators may need to clarify how existing advertising and labeling requirements apply when product information is communicated through AI assistants.
These questions become more relevant as shopping assistants acquire greater transactional capabilities. Research published in 2026 has examined ecommerce models in which autonomous agents monitor markets and make purchasing decisions on behalf of consumers, while Amazon has already introduced features that allow its shopping technology to make purchases when products reach customer-defined prices.
As shopping systems assume more responsibility for product research and selection, manufacturers will need to consider how accurately their product attributes can be retrieved and evaluated by machines. The issue is not limited to whether an AI assistant can generate a useful answer, since incomplete or inconsistent product information can affect whether a product appears in the customer’s consideration set in the first place.
Made in USA provides a useful case study because country of origin carries commercial and regulatory implications, but similar questions are likely to arise around sustainability claims, component sourcing, repairability, certifications and other product characteristics that require reliable supporting information.
For manufacturers, the development of AI shopping assistants represents a change in the mechanics of ecommerce product discovery rather than a replacement for existing sales and marketing practices. Search visibility, product content and marketplace performance will remain relevant, but manufacturers may also need to treat accurate, structured and verifiable product data as part of their approach to digital distribution as AI becomes more involved in deciding which products consumers are shown.
Sources:
Forbes