Costco's AI recommendation carousels: the '$470 million' and '3x conversion' figures, in the company's own words

The problem

Costco Wholesale, the membership warehouse retailer, frames AI and personalization as levers on the part of its business that grows fastest: e-commerce. On its FY2026 earnings calls the company grouped its on-site personalization work and its AI-search results together as digital enhancements and put specific business figures against them [source]. The question a buyer would ask is whether a recommendation engine actually moves e-commerce sales, and by how much.

What was built

Costco’s chief financial officer, Gary Millerchip, described two distinct systems on the calls. The first is a set of “personalized product recommendation carousels,” an on-site recommendation and personalization system, sitting alongside “newly modernized product display pages” [source]. The second is AI search on the Costco site and app, which the company reports on separately from the carousels ([source]). CEO Ron Vachris placed both inside a broader AI strategy, saying the company is “working closely with the leading AI companies to ensure our values will be visible to existing and potential future Costco members” [source].

The outcome

Costco put a headline number on the carousels. On the Q2 FY2026 call the CFO said, “In Q2, our personalized product recommendation carousels drove over $470,000,000 of ecommerce sales” [source], a figure independent trade press reported as “more than $470 million in e-commerce sales during the quarter” ([source]). On the Q3 FY2026 call he added that “our personalized product recommendation carousels delivered conversion rates 3x better than our typical conversion rates” [source], which Digital Commerce 360 reported as “three times better than its typical conversion rates” ([source]).

On AI search, the CFO said, “While the volume of traffic generated from AI search is still low, we saw triple digit growth in Q3, and this activity had the highest conversion rate of all traffic coming to our site” [source], a characterization independent trade press repeated ([source]). He also reported that “in digital, we saw strong member engagement in Q3, with site and app traffic up 37%” [source], which Digital Commerce 360 reported as traffic that “increased 37% year over year” ([source]).

One Q3 figure is deliberately left off this page. Costco reported a Q3 e-commerce-sales figure for the carousels, but the sources do not reconcile: the Motley Fool transcript prints “just under $5 billion” while Digital Commerce 360 and Customer Experience Dive both report the figure as “just under half a billion” / “nearly $500 million,” which is the reading consistent with the $470 million Q2 base [source]. Because the two figures cannot both be right and no primary filing settles it here, this case file treats the Q3 dollar amount as unresolved and does not rely on it ([source]).

The load-bearing caveat is who is doing the counting. Every figure here is Costco’s own, stated by its CFO on an earnings call, and the $470 million is an attribution the company makes to its own carousels rather than an independently measured causal lift [source]. The weakest load-bearing sources on this page are the two trade publications, Customer Experience Dive and Digital Commerce 360, and they are aggregators of the earnings call rather than independent measurement, so they repeat Costco’s numbers without testing them ([source]). Attributing a share of sales to a recommendation engine also cannot separate the engine’s contribution from members who would have bought anyway, and nothing in the disclosure isolates that.

Why this caps below green

Every load-bearing figure on this page traces to Costco. An earnings call is a regulated setting, which is why these claims are documented rather than marketing, but a company reporting a favourable attribution about its own systems is not the same as an independent audit of them. Under TIN’s rule, green never rests on the subject’s own numbers, and there is no non-Costco measurement to lean on here. The honest ceiling is corroborated.

Related case files

• Klarna’s AI assistant and its customer-service figures — the closest parallel: a company’s own AI system reported through the company’s own business metrics, the same first-party-attribution shape as Costco’s carousels.
• The DOJ/SEC case against the Nate AI shopping app — an AI shopping-assistant claim seen from the other end, where prosecutors found the automation figure was fabricated, showing why first-party attribution alone cannot earn green.
• The FTC’s Automators AI e-commerce case — another AI-in-retail claim that did not survive independent scrutiny, ending in a judgment and lifetime ban.