Two U.S. senators asked the Federal Trade Commission on September 17, 2026 to investigate how Amazon and Walmart’s AI shopping assistants handle “Made in USA” products. The request is not an enforcement action, but it puts country-of-origin data directly inside the debate over how marketplaces recommend products.

Reuters reports that Senators Tammy Baldwin and Rick Scott based their letter on research from Columbia’s Center for Law and Economy. The study alleges that Amazon’s Alexa for Shopping and Walmart’s Sparky can identify suspicious origin claims but do not reliably flag them or help shoppers find American-made alternatives.

Amazon told Reuters that any suggestion it intentionally withholds country-of-origin information is wrong. It said Alexa currently sends shoppers to product detail pages for origin information while Amazon works to improve accuracy. Walmart did not immediately respond, and the FTC has not said it will open an investigation.

This is an allegation about AI behavior, not a new seller rule

The careful version of this story matters. Two senators want an investigation. A research group produced findings. Neither fact proves that Amazon or Walmart violated the law, and sellers have not received a new origin-data requirement.

The underlying compliance standard is real. Under the FTC’s Made in USA guidance, an unqualified U.S.-origin claim generally requires the product to be “all or virtually all” made in the United States. Qualified claims such as “Assembled in USA from imported parts” need to describe the actual basis accurately.

Marketplace enforcement has already been on the FTC’s radar. The agency warned Amazon and Walmart in July 2025 about Made in USA claims from third-party sellers. The September 17 letter asks whether the platforms’ own AI tools should now do more with the information already in their catalogs.

AI assistants create a second product-discovery layer

Sellers normally optimize for category placement, search terms, conversion and the Featured Offer. A conversational assistant can bypass much of that path by answering a shopper’s question with a short list of products.

That pattern is visible in how Instacart’s AI builds carts and chooses lower-cost substitutes. Once an assistant assembles the shortlist, a shopper may never see the category page where your listing was designed to compete.

Country of origin can therefore become a recommendation input. If the field is missing, inconsistent or buried in an image, the assistant may not use it. If a competitor makes an unsupported claim that the assistant accepts, that competitor may receive traffic it should not have.

Amazon’s response also exposes a practical limitation. Sending the customer to a product page is different from answering the question. The page may contain conflicting text across the title, bullets, attributes, packaging images and seller-supplied content. An AI system that prioritizes accuracy may avoid a direct answer when those records disagree.

Audit the claim before testing the chatbot

Start with your evidence, not the marketing copy. Map where the product was manufactured, where its important components originated and where final processing occurred. Preserve bills of materials, factory declarations, purchase orders and any testing or production records that support the claim.

Then compare every public version of the claim. Check packaging, inserts, your DTC site, Amazon and Walmart attributes, titles, bullets, enhanced content and advertising. “Designed in USA,” “Assembled in USA” and “Made in USA” are not interchangeable phrases.

Do the same review for implied claims. A U.S. flag, map, factory image or wording such as “American quality” can contribute to the overall impression even when the exact phrase “Made in USA” never appears. If your evidence only supports a qualified claim, use one that is clear near the claim rather than hiding the qualification in fine print.

After the listing is consistent, test the shopping assistants with neutral prompts. Ask where the product is made, request American-made options in the category and compare the response with the product detail page. Save the prompt, answer, date, marketplace and ASIN. This is monitoring evidence, not proof of how every shopper will be treated.

Treat false competitor claims as an evidence problem

Do not report a competitor simply because its price looks too low for U.S. production. Capture the exact claim and gather credible evidence showing why it may be inaccurate. A platform is more likely to act on a specific mismatch than a general accusation.

The bigger issue is that Amazon keeps more control over product data even while AI becomes a traffic gateway, a strategy visible in Amazon’s approach to advertising on ChatGPT without broadly sharing its catalog. Sellers depend on how Amazon structures and interprets that controlled data.

For now, no seller should rewrite listings around one research report. Make the origin record accurate, consistent and defensible. The next material development would be an FTC investigation, a platform change to origin attributes or evidence that country-of-origin data directly affects AI recommendations. Those are the points where this moves from a warning signal to an operational policy story.