Holiday 2026 is the first season AI shopping agents are expected to really spend money
Retail data now shows AI shopping agents moving from demos to real purchases, though shoppers and brands are not fully ready to let them check out alone.
The short version
- eMarketer forecasts $1.44 billion in U.S. e-commerce sales completed end to end by agents this year.
- Adobe says September visitors sent by AI tools bought at a rate 90% higher than other visitors.
- Okta found only 31% of consumers likely to let an agent buy without reviewing first.
- Gartner: 61% of senior marketers adapted content for AI, but 30% built workflows for agent transactions.
- Walmart lets shoppers cap an agent's spending at $50, $100 or $200.
Why it matters
If software starts comparing products and paying for them, stores have to sell to a machine as well as a person. Shoppers get a new convenience, and a new decision about how much spending power to hand over. The charts suggest the technology is moving faster than trust.
From demo to checkout
For a while, AI shopping agents, meaning software that researches products and buys on your behalf, mostly lived in product demos. As Modern Retail reported in a chart-led feature first published by its sister site Digiday, this holiday season looks like the first where the agent actually pays, making the fourth quarter the agentic commerce quarter.
The week's news backs that up. OpenAI launched a wallet inside ChatGPT meant to push agent purchases through retailers' own checkouts, along with a playbook for businesses that want to prepare. Google started rolling out a checkout built on its UCP system in Canada, through AI Mode and Gemini.
Not everyone is opening the door. Amazon blocked Meta's Muse AI agent from its retail site. Walmart, meanwhile, lets customers cap what an agent can spend at $50, $100 or $200. Bain estimates roughly one in four online shoppers plan to begin their holiday shopping on an AI platform.
The money is small, but moving
The first forecast number is modest. eMarketer expects U.S. retail e-commerce sales completed fully by agents, from request to payment, to reach $1.44 billion this year. It projects growth of 10.6% to $1.59 billion in 2027, then 11% to $1.76 billion in 2028.
The leap comes after that: the same forecast reaches $2.17 trillion by 2030. That is a very long way from today's figure, and forecasts that far out deserve caution. Still, it shows how seriously the industry is planning for this.
Brands prepare to be found, not bought from
Marketers are reacting, but mostly on the discovery side. In Gartner's 2026 Marketing Strategy and Engagement Survey, 61% of senior marketing leaders said they had adapted their content and search work so machines can interpret it. That includes SEO and AEO, which are practices for getting found by search engines and AI answer tools.
Another 42% are watching for and correcting false statements about their brands in AI-written answers, and 39% have tailored product details, prices or offers for machine decision-making.
Fewer have gone further. About 30% have built interfaces or workflows for agent transactions, and 32% treat machines and agents as a separate audience next to human customers. In short, brands are working to be recommended before they are ready to be bought from.
Shoppers want to keep a hand on the wheel
Okta, an identity and access management company, asked consumers in its Customer Identity Trends Report 2026 whether they would let an agent finish a purchase without reviewing it first. Only 31% said they were very or rather likely to. Some 67% said they were unlikely, and 41% said not at all.
Early adopters of the technology look different. Among them, 62% would let an agent complete a purchase unchecked, and 42% said they were very likely to. The gap points to trust, not engineering, as the hard part.
Yet agents already shape where people shop. Adobe Digital Insights found that in September, visitors who reached U.S. online retail sites through AI tools bought at a rate 90% higher than visitors from other channels. That is the best level Adobe has recorded. In August 2025, AI-referred visitors bought at a rate 9% lower than others.
Amazon's ad business is the one to watch
Agents could change how retailers earn money, not only how they sell. Enders Analysis puts Amazon's third quarter of 2026 at $68.5 billion in net sales from its own online stores. Seller services for outside merchants brought in $46.5 billion, and advertising $20 billion.
The advertising line is the one at risk. If shoppers lean on an agent to pick products instead of browsing sponsored listings, fewer people may ever see a paid placement. Amazon's warehouses and delivery network could still serve orders that start somewhere else, but the ad revenue depends on people looking at its pages.
What to watch
- Whether more shoppers become comfortable letting an agent pay without reviewing the order first.
- How Amazon responds as agents from other companies try to shop on its site.
- Whether brands move from being discoverable by AI to being ready for agent-led transactions.
Try it
If you sell online, check that titles, prices, sizes, stock status and return policies match everywhere they appear. If you shop, test an agent on a small purchase and look for a spending cap first.
Sources and links
First reported by Modern Retail. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.