Retailers spent the first half of 2026 declaring victory on AI shopping agents. Walmart, Amazon and Shopify all told investors their assistants are driving bigger orders. eBay spent the same stretch writing agents out of its terms of service entirely. Both reactions cannot be evidence of the same thing working. Read closely, the retailers' own numbers do not prove what they are being used to claim, and the platform split over whether to let agents in at all is the more honest signal of where things stand.
What the retailers reported
Amazon merged its Rufus and Alexa+ shopping tools into a single assistant, Alexa for Shopping, and CEO Andy Jassy told investors on the company's July 30 earnings call that more than 350 million customers had used it over the prior 12 months, with active users nearly doubling year over year and total interactions up more than fivefold. Customers who use it, Jassy said, spend more than 40% more per order than those who do not.
Walmart reported something similar two months earlier. U.S. CEO David Guggina told Modern Retail in May that customers using the company's Sparky assistant have an average order value about 35% higher than non-users, and that roughly half of Walmart's app users have tried it. Shopify reported that AI-referred traffic and orders tripled year over year, with AI already reaching purchases outside the categories shoppers buy on repeat, according to PYMNTS' review of the quarter's retail earnings.
These are real, disclosed figures from earnings calls and named executives, not estimates. The problem is what they are being asked to prove.
The gap between correlation and the claim
None of these numbers isolate what the assistant caused. Users Know, a product research publication, pointed out in its review of Walmart's disclosure that the company's most engaged app users are the ones most likely to notice and try a new feature like Sparky in the first place, and that this same group was probably already spending more before the tool existed. A higher average order value among Sparky users could just as easily describe who tries Sparky as what Sparky does to their spending.
Walmart and Amazon are not alone in this gap. PYMNTS' own reporting on the quarter noted that AI-driven spending lift has mostly shown up in categories shoppers already buy on repeat, and that whether it holds for a purchase the shopper was not already planning remains unanswered. TJX, which sells almost entirely through in-store browsing, posted 4% comparable sales growth with no AI framing at all, a reminder that the lift is concentrated in retail formats built for it, not universal.
None of this means the assistants are not doing anything. It means retailers are citing selection effects as proof of causation, and no earnings call so far has controlled for the difference.
The platforms saying no
While Walmart and Amazon were building agents into their storefronts, eBay was writing them out of its own. New terms of service that took effect February 20 bar "buy-for-me agents, LLM-driven bots, or any end-to-end flow that attempts to place orders without human review" from operating on the platform without eBay's permission. The stated concern is attribution: an agent that completes a purchase on a shopper's behalf can break the tracking that affiliate and advertising revenue depend on, and eBay wants commissions tied to a session it can still see.
Google took the opposite position at NRF in January, announcing the Universal Commerce Protocol, an open standard for agentic commerce co-developed with Shopify and backed by more than 60 companies including Mastercard, Visa, PayPal, Target and Best Buy. Google's agentic checkout, which lets a shopper's AI agent complete a purchase through a "Buy for me" button, has been live with select U.S. merchants since November.
So the industry has split. Some retailers are building agent-native checkout into the storefront itself. Others are citing engagement numbers they have not yet shown caused anything. eBay decided the safer move was to keep agents out of the transaction entirely. That split, not the spending percentages, is the fact worth watching this year.
What would actually settle it
A defensible causal claim would compare spending by the same customers before and after they started using an assistant, or run a controlled test between AI-assisted and unassisted shopping at similar starting points. Neither Walmart nor Amazon has published that comparison. Until one does, "customers who use it spend more" is a description of who adopts the feature first, not proof the feature is what moves the number.
Sources
- About Amazon, "Amazon CEO Andy Jassy: Stores growth, AI shopping, Q2 2026 earnings" - https://www.aboutamazon.com/news/company-news/amazon-ceo-andy-jassy-stores-growth-ai-shopping-q2-2026-earnings
- Retail Dive, coverage of Amazon Alexa for Shopping agentic AI adoption - https://www.retaildive.com/news/amazon-alexa-for-shopping-agentic-ai-adoption/826806/
- Modern Retail, "Walmart says AI users build 35% bigger baskets than others" - https://www.modernretail.co/technology/walmart-says-ai-users-build-35-bigger-baskets-than-others/
- Users Know, "Walmart says their AI investment is working. The metrics they presented don't actually show that." - https://www.usersknow.com/blog/2026/5/26/walmart-says-their-ai-investment-is-working-the-metrics-they-presented-dont-actually-show-that
- PYMNTS, "Retailers report AI-driven sales, bigger baskets in Q2 earnings" - https://www.pymnts.com/news/artificial-intelligence/2026/retailers-report-ai-driven-sales-bigger-baskets-q2-earnings/
- PYMNTS, "eBay blocks use of AI buy-for-me agents" - https://www.pymnts.com/artificial-intelligence-2/2026/ebay-blocks-use-of-ai-buy-for-me-agents/
- Google Blog, "Agentic commerce AI tools protocol for retailers, platforms" - https://blog.google/products/ads-commerce/agentic-commerce-ai-tools-protocol-retailers-platforms/
- Google Developers Blog, "Under the hood: Universal Commerce Protocol (UCP)" - https://developers.googleblog.com/under-the-hood-universal-commerce-protocol-ucp/
- Google Blog, "Agentic checkout for the holidays: AI shopping" - https://blog.google/products-and-platforms/products/shopping/agentic-checkout-holiday-ai-shopping/