AI assistants are on course to originate an estimated $9.0 billion of U.S. retail purchases in 2026, growing at a 52.7% CAGR to $74.7 billion by 2031, and they mostly route shoppers to retailers’ own checkouts rather than buy for them.

On October 15, 2026, Google launched its Universal Cart across U.S. Search and the Gemini app, with checkout through the Universal Commerce Protocol and the retailer remaining the seller of record.

AI shopping assistants mostly route U.S. shoppers to retailers’ own checkouts rather than buy for them. The contest moves to visibility inside the AI answer, which Google now sells, a cost the sources reviewed for this report do not yet set against the sales it originates.

Those assistants are on course to originate an estimated $9.0 billion of U.S. retail purchases in 2026, a full-year estimate. This report forecasts those purchases to grow at a 52.7% CAGR to $74.7 billion by 2031.

The report sizes the market by who submits the order and by the door it flows through: platform protocol, retailer-owned agent or walled garden. Its 80 pages carry three scenarios, a year-by-year forecast, a two-stage path to 2036, 14 charts and 12 data tables. Fourteen companies are profiled, including Google, Amazon, Walmart, Shopify, OpenAI and Stripe.

Report Highlights:

• Who Submits the Order: AI assistants mostly route U.S. shoppers to retailers’ own checkouts rather than buy for them. Of an estimated $9.0 billion in AI-originated U.S. retail sales for full-year 2026, agents themselves are estimated to submit a small minority, and well under half even at the top of the report’s bound. Walmart reports that partner assistants such as ChatGPT and Gemini mostly deep-link shoppers out to its own site.
• Market Size and Forecast: AI-originated U.S. retail sales grow from $9.0 billion in 2026 to $74.7 billion by 2031, a 52.7% CAGR, with the report’s three scenarios bracketing that path. In the base case, that is faster than U.S. e-commerce grew after 1999 from almost the same starting share (then 0.6% of total retail; this market’s 0.65% is of e-commerce), reaching 2.2% of retail in five years. On a two-stage path, growth slows after 2031, and the report carries the forecast to 2036.
• The Faster-Growing Lines and Doors: In the report’s base-case allocation, agent-submitted orders are the faster-growing of the two lines as protocol checkout matures, well ahead of agent-routed orders. Of the three doors, retailer-owned agents, led by Walmart’s Sparky, grow fastest in that allocation, narrowly ahead of the walled garden and well ahead of platform protocol. The report gives each line and door its value and growth rate through 2031.
• Where the Platforms Earn: Platforms now earn on visibility, not on the checkout: Google’s merchant terms for UCP checkout keep the retailer as merchant of record and name no per-order fee. Meanwhile, Google sells ads inside AI Mode answers and has added native UCP checkout to its Direct Offers. Of the seven competing reports whose listings were read for this study, the catalogue reports size vendor software revenue, such as MarketsandMarkets’ estimate of about $1.06 billion in 2025, global. Among the others, yStats argues that control of interfaces decides position, and ResearchFarm reports a consensus that agentic AI mainly threatens retail media. The seven listings show no split of U.S. purchase value by who submits the order.
• Delegation and Conversion: Delegation remains the least proven part of the market, and conversion depends on the door. Shopping agents built on 2025 state-of-the-art models such as GPT-4.1 succeed on fewer than half of realistic benchmark tasks. In a peer-reviewed study of 973 websites, ChatGPT referrals converted below Google’s paid and organic search. Yet Walmart reports strong conversion on the shoppers partner assistants send to its site. Shopify reports that AI searches using its product catalog convert at twice the rate of searches using scraped data.

This report will provide answers to the following questions:

• How large is the U.S. agentic commerce market in 2026, and how fast will it grow to 2031?
• What share of AI-originated U.S. retail sales do agents themselves submit, and how much do assistants route to retailers’ own checkouts?
• What did Google’s Universal Cart change when it launched on October 15, 2026, and what does the retailer keep as seller of record?
• Which door grows fastest through 2031: platform protocol, retailer-owned agent or walled garden?
• Where do the AI platforms earn: on visibility inside the AI answer, on checkout or on payments?
• How does the forecast compare with U.S. e-commerce’s growth after 1999, and how do the three scenarios bracket it?
• How are the leading platforms, retailers and payment networks positioned, and what do their disclosures allow?
• How reliable are shopping agents today, and how does conversion differ by door?

Companies Covered:

Google, Amazon, Walmart, Shopify, OpenAI, Meta, PayPal, Stripe, Visa, Mastercard, American Express, Newegg, Riskified, Crossmint

Report Specifications:

80 pages · 14 charts and figures · 12 data tables · 14 company profiles · 5-year forecast horizon, to 2031 · 3 scenarios · 2 segmentation views

Pricing:

• Single User License – $9,950
• Site License – $10,950
• Enterprise License – $11,950

To place an order, contact david.pringmill@policy2050.com