Ecommerce 2026: How AI Agents Rewired $262B in Sales

By Theo Werner 2026-07-18 15 min read

AI agents and generative AI tools influenced roughly $262 billion in global online sales during the 2025 holiday season, about 20% of all orders, and that single figure explains why ecommerce 2026 no longer resembles the ecommerce of 2023. Salesforce data shows referral traffic from generative AI platforms climbed 693% year over year in November and December 2025, and that traffic converted 31% better than other sources. The channel brands spent a decade optimizing for human eyeballs is now being read, filtered, and checked out by software. This analysis breaks down the agentic shift, the payment rails behind it, the regulation catching up, and what growth teams should change first.

What ecommerce 2026 actually looks like

Strip away the noise and the 2026 market is defined by a few hard numbers. Global ecommerce is on track for roughly $6.8 trillion in sales this year according to eMarketer, and online now accounts for about 21% of total retail worldwide. In the United States, U.S. Census Bureau figures put retail ecommerce at $326.7 billion in the first quarter of 2026, or 16.4% of all retail spending, up from 15.9% a year earlier. Headline growth has cooled to high single digits. The real story of 2026 is not how much people buy online, it is how they buy.

Mobile is the substrate. Around 59% of online retail sales now happen on phones, while mobile drives close to 78% of all ecommerce traffic. That gap between traffic share and revenue share is where a lot of margin leaks, and it is why conversion rate optimization on small screens remains the least glamorous and most profitable work most brands ignore.

Layered on top of mobile is the defining shift of the year: agent mediation. Shoppers increasingly start with a chat interface rather than a search box, and software increasingly finishes the transaction. The 693% jump in AI referral traffic was not a holiday blip. Adobe reported that AI-driven traffic to US retail sites rose 393% year over year in the first quarter of 2026, with agentic shoppers spending more per visit than human-only sessions. Ecommerce has become business-critical infrastructure that has to serve two very different customers at once: the person and the agent acting on their behalf. Any serious ecommerce growth strategy in 2026 has to plan for both.

The agentic checkout stack: OpenAI, Google, Visa, Mastercard

Behind the agentic shopping surge sits a stack of new protocols and payment rails that mostly did not exist eighteen months ago. In the span of a single year, four of the largest technology and payments companies shipped competing standards for letting software buy on a human's behalf, and the fight over which one wins is now one of the defining commercial contests of 2026.

The opening move came in spring 2025. Mastercard and Visa unveiled agentic capabilities within days of each other. Mastercard launched Agent Pay on April 29, 2025, built on agentic tokens and Payment Passkeys, and has since run live authenticated agentic transactions in Hong Kong and Thailand. Visa launched Intelligent Commerce on April 30, 2025, giving developers a single integration for agent identity, spending controls, and checkout. Then in September, the platform giants arrived. Stripe and OpenAI released the Agentic Commerce Protocol on September 29, 2025, powering an Instant Checkout experience inside ChatGPT. Two weeks earlier, Google announced the Agent Payments Protocol on September 16, 2025, with more than 60 launch partners including Mastercard, American Express, PayPal, Coinbase, Adyen, and Salesforce.

These standards are not interchangeable. AP2 uses signed Intent, Cart, and Payment mandates carried as verifiable credentials and treats stablecoins as first-class payment rails alongside cards. ACP is a lighter open standard that a Stripe merchant can enable in about one line of code. The table below maps the players a growth team needs to know.

The strategic read is simple. Standards are consolidating faster than most retailers can integrate, and by June 2026 Forbes was already describing Visa, Mastercard, and Coinbase as fighting over how AI agents pay. Betting on a single rail is risky. Building a clean, machine-readable product feed that any of these protocols can consume is the safer investment.

Why OpenAI pulled back from Instant Checkout in March 2026

The most instructive news event of the year was a retreat, not a launch. On March 5, 2026, less than six months after the fanfare, OpenAI shifted away from Instant Checkout as a direct-purchase feature and pointed transactions toward merchant-specific apps inside ChatGPT instead. Only about 30 Shopify merchants had integrated the original flow. Retailers including Target, DoorDash, and Instacart launched dedicated ChatGPT apps to handle purchasing inside their own experiences.

The reasons matter more than the headline. Shopify president Harley Finkelstein noted on March 3, 2026 that checkout is not just a payment. It bundles subscriptions, inventory management, shipping, taxes, and merchandising options that demand constant updates, and stripping all of that into a generic protocol turned out to be harder than a demo suggests. An OpenAI spokesperson framed the pivot plainly: "We're prioritizing making ChatGPT search and product discovery great, with ACP serving as the infrastructure that connects users to merchants across the full shopping journey."

For growth and SEO professionals, the lesson is worth more than any protocol spec. Agentic discovery is scaling fast and works. Agentic checkout, the last click where money moves, keeps running into the operational reality of running a store. The near-term winners are brands that make themselves easy to discover and compare inside agents, then hand shoppers into a checkout they fully control. Do not rebuild your entire payment flow around one platform that may change strategy in a quarter. Do make your catalog legible to every agent that might surface it. The protocol layer is still settling, and betting the store on any single one is premature.

Retailer-built agents are beating the platforms

While the platforms iterate, retailers that built their own agents are posting the clearest returns. Walmart launched Sparky in June 2025 as a goal-based assistant designed to replace keyword search, letting a shopper say "plan a cookout for eight" and getting a full cart back. Sparky now runs inside the Walmart app, ChatGPT, and Google Gemini, and by May 2026 Walmart credited Sparky with an average order value roughly 35% higher for users than non-users. Walmart has since opened advertising placements inside Sparky, turning the assistant into a retail-media surface as well as a shopping tool.

Amazon is racing in parallel. Its Buy for Me feature can shop and purchase from external retailers without human review, and in 2026 Amazon began offering the underlying technology to outside retailers through an AWS Agentic Shopping Assistant built on the same stack that powers Rufus and Alexa shopping. Perplexity shipped free agentic checkout to all US users. Salesforce reported that retailers running their own shopper agents grew sales 59% faster than those still on the sidelines. The mid-2026 agent landscape has settled into a short list of serious players:

• Walmart Sparky, embedded across the app, ChatGPT, and Gemini, with a documented 35% AOV lift.
• Amazon Rufus and Buy for Me, now packaged for third-party retailers via AWS.
• OpenAI ChatGPT, pivoting from Instant Checkout to merchant apps and product discovery.
• Perplexity, with free agentic checkout for every US shopper.
• Google Gemini, distributing agent commerce through AP2 and its partner network.

Caila Schwartz, Director of Consumer Insights at Salesforce, summarized the 2025 holiday result this way.

The 2025 holiday season marked a definitive shift to a new era of agentic shopping. While shoppers remained resilient in the face of higher prices, the real story was how retailers leaned on AI and agents to navigate the holiday rush. Agents did not just drive $262 billion in sales through high-intent discovery; they became the operational heroes of the season.

The takeaway for brands without Walmart's engineering budget is that owning even a narrow, category-specific agent now beats waiting for a platform to send qualified traffic on favorable terms.

What shoppers actually want from AI, and what they do not

Underneath the adoption curve sits a tension that growth teams keep misreading. Consumers are enthusiastic about AI as a research and comparison tool and far more cautious about handing over the final decision. The DHL eCommerce Trends Report 2026, published June 2, 2026 and based on 29,000 shoppers and 5,800 businesses across 29 countries, found that 29% of shoppers would be happy to hand shopping control to AI within five years, rising to 33% of Gen Z and 36% of millennials. Other surveys land higher or lower depending on how the question is framed, which is exactly the point.

Gartner's May 2026 survey is the most sobering read. Willingness to let AI make the actual purchase decision topped out at 11% even in low-stakes categories like personal care, while 31% would let AI narrow choices for household supplies. Trust is fragile in another way too: a separate April 2026 consumer poll found 75% of Americans would lose trust in AI shopping if the results were sponsored. Kate Muhl, VP Analyst at Gartner, drew the line clearly.

Consumers are not looking to outsource shopping decisions to AI. They want AI to help them find better information, compare prices, identify deals and narrow choices, while keeping final decision-making control for themselves.

The strategic implication is that discovery and consideration are where agents create value in 2026, and where brands should invest first. The pixel-perfect autonomous checkout will come, but the money this year is in being the option an agent recommends and a human confidently approves.

Optimizing the digital shelf for machines, not just humans

If agents now sit between your catalog and your customer, then your product data is your storefront. Adobe's own research carried an uncomfortable finding for retailers: AI traffic is surging, but most retail sites are not machine-readable enough for agents to parse cleanly, leaving brands invisible in AI-driven results even as that traffic converts at eight times the rate of social. The digital shelf has a second audience now, and it does not squint at pretty product photography. It reads structured data.

This is where classic SEO meets the newer disciplines of answer engine optimization and search visibility. The tactics that make a product legible to a shopping agent overlap heavily with technical SEO fundamentals, but the stakes are higher because an agent will silently skip anything it cannot parse. A practical 2026 checklist looks like this:

• Mark up every product with complete schema.org Product data, including GTINs, price, and real-time availability.
• Maintain a clean, current product feed that ACP, AP2, and retailer agents can ingest without guesswork.
• Expose accurate stock, shipping, tax, and return information as structured fields, not buried prose.
• Treat verified reviews and user-generated content as ranking fuel, since agents weight authentic signals over polished copy.
• Answer real buyer questions in structured Q and A blocks that an agent can quote directly.
• Monitor how your top products render inside ChatGPT, Perplexity, Gemini, and Google AI Overviews, and fix gaps weekly.

The mental model shift is the hard part. For fifteen years, ecommerce teams optimized pages for a human scanning a grid. In 2026 they also optimize records for a model comparing options across a thousand merchants in a second. Polish still matters for the human who approves the cart, but structure decides whether the agent ever surfaces you at all. Brands that treat feed quality as an engineering priority rather than a marketing afterthought are the ones agents recommend.

Livestream and social commerce cross $68 billion

While agents dominate the discourse, a more human channel quietly crossed a milestone. US livestream commerce sales are projected to reach roughly $68 billion in 2026, up about 36% from around $50 billion, and now account for more than 5% of total digital commerce. The engine is social. Forbes reported that TikTok Shop sellers sit behind an $87 billion social commerce surge, and TikTok Shop alone is projected to hit $23.4 billion in US sales in 2026, a 48% year-over-year jump that makes it a larger US ecommerce business than Target, Costco, Best Buy, or Kroger.

The conversion math is what keeps merchandising teams interested. Live shopping events can convert at up to 30%, against the 2% to 3% typical of standard ecommerce, because a real person is demonstrating the product and answering objections in real time. The ceiling is adoption, not performance: only about 12% of US shoppers have bought through a livestream so far, with another 12% saying they plan to try. That leaves a long runway for beauty, apparel, and lifestyle brands willing to build a repeatable live-selling motion rather than one-off events.

For growth teams, livestream and social commerce sit at the intersection of content, creator partnerships, and paid advertising, and the retail-media angle is compounding. Walmart putting ads inside Sparky and TikTok Shop scaling its ad load both point the same direction: the surfaces where shoppers discover products are becoming ad-funded, whether those surfaces are livestreams or AI agents. The brands winning here treat live selling as an always-on channel with its own creative calendar, not a novelty, and they measure it on incremental revenue rather than concurrent viewers.

Sustainability becomes table stakes, and regulation catches up

The second half of the DHL report reframes sustainability from a premium feature into a baseline expectation. Sustainable logistics, once a differentiator, is now expected as standard by 42% of consumers within five years. More concretely, 35% of shoppers said they have abandoned an online purchase over a lack of sustainability credentials, and 32% of businesses confirmed customers have abandoned carts for that reason. Second-hand and recommerce are going mainstream at the same time. Sustainability is no longer a marketing story, it is a conversion variable.

Regulation is arriving, though not on the timeline many briefs claim. It is worth correcting a common error: the EU's Digital Product Passport is not a blanket January 1, 2026 mandate forcing every fashion and furniture brand to publish supply-chain data. The reality is more staggered. The European Commission's ESPR Working Plan for 2025 to 2030 was adopted April 16, 2025. Batteries are the first product group with a mandatory passport, taking effect in February 2027 under the separate EU Battery Regulation. The textiles delegated act is only expected around 2027, and because delegated acts cannot apply earlier than 18 months after entry into force, mandatory textile compliance is realistically no earlier than 2028. The EU is expected to stand up its central digital registry around July 2026.

That correction matters because it changes what a brand does now. There is no fire drill for a January 2026 deadline that does not exist. There is a clear case for starting a product-data inventory in 2026 so that when your category's deadline lands, the passport is a data-plumbing exercise rather than a scramble. Pablo Ciano, CEO of DHL eCommerce, tied the AI and trust threads together at the report's launch.

The ability to understand and respond to customer needs has always defined success, but our new eCommerce trend report shows that AI is now redefining that advantage at hyperspeed. Consumers can identify the best offer in milliseconds, and retailers can gain insights that allow them to instantly capitalize on changing demand. In this new era, the winners will be those who move fastest, and translate that speed into superior customer experiences.

B2B ecommerce and the efficiency reset

The consumer headlines obscure where the largest dollars actually move. Global B2B ecommerce is valued at roughly $36 trillion in 2026 and is growing at about a 14.5% compound annual rate through 2030, according to International Trade Administration and industry data. Amazon Business alone now serves more than 6 million customers and generates around $35 billion in annual sales, up from $1 billion in 2016 and $25 billion in 2020, and it signed 41% more enterprise accounts in 2025 after extending multi-channel fulfillment. The same agentic dynamics reshaping consumer checkout are digitizing complex procurement workflows for manufacturers and distributors.

B2B buyers are also consumers in their off-hours, and they now expect consumer-grade experiences: self-serve catalogs, real-time pricing, mobile approval flows, and increasingly, agents that can reorder or source against a target price. The strategic frame for 2026 that keeps recurring across analyst commentary is an efficiency reset, where reliability of operations supersedes surface innovation. Fast, accurate, and consistent beats novel. That means loyalty programs treated as data engines rather than points ledgers, verified reviews and user-generated content treated as the credible trust signal in an AI-flooded web, and composable or headless architectures adopted to unify inventory and pricing across every channel a human or an agent might touch.

For both B2B and DTC operators, the operational bar has risen because agents punish inconsistency ruthlessly. A price that is stale in your feed, a stock count that is wrong, or a shipping estimate that does not match reality will get your product filtered out before a human ever sees it. The unglamorous work of data hygiene has become a growth lever. Teams that win in 2026 spend less on surface novelty and more on making sure every system of record agrees with every other one, because the machines shopping on their behalf assume it already does.

Five predictions for 2026 and 2027

The trajectory of the last twelve months supports a handful of specific, testable calls for the next eighteen. These are the shifts a growth team should be positioning for now, not reacting to later.

• Payment-standard consolidation. By the end of 2026, the field of competing agent-payment protocols will narrow from four-plus toward two dominant stacks, with AP2 and ACP absorbing most of the merchant integrations and card networks bridging both.
• Retailer agents outpace platform agents. Expect more brands to report Sparky-style average order value lifts through 2027, and expect white-label tools like the AWS Agentic Shopping Assistant to reach mid-market retailers who cannot build their own.
• AI referral traffic becomes a named channel. After a 693% holiday jump and a 393% Q1 2026 rise, generative-AI referrals will appear as a standard line item in retail analytics dashboards by the fourth quarter of 2026, not a rounding error inside direct traffic.
• Autonomous checkout stays a minority behavior. Given Gartner's 11% ceiling and the 75% sponsored-results trust cliff, agents will win discovery and comparison far faster than final purchase authority through 2027.
• Product passports pull brands in early. The February 2027 battery passport deadline will drag textiles and electronics brands into voluntary DPP pilots during 2026, because retailers want one product-data model, not three separate compliance projects.

None of these require a leap of faith. Each one is an extrapolation of a trend already visible in the 2025 and early 2026 data, which is exactly why they are worth building against rather than debating.

How to get started: what to do Monday morning

The gap between reading about agentic commerce and being ready for it is a to-do list, not a transformation program. Here is what a growth, SEO, or ecommerce lead should action this week.

• Run an agent visibility check. Search your top 50 products inside ChatGPT, Perplexity, Gemini, and Google AI Overviews. If an agent cannot read your price, availability, and specs, you are invisible to it, and a structured AI visibility audit will show you exactly where the gaps are.
• Ship a clean product feed. Complete schema.org markup, GTINs, and real-time stock are the price of entry for ACP and AP2 surfaces.
• Decide your agent posture. Build a narrow category agent, plug into ChatGPT apps and Sparky-style retail channels, or do both, but decide deliberately rather than by default.
• Instrument AI referral traffic as its own channel so you can see the 693% trend in your own numbers, not just headlines.
• Fix mobile checkout friction first, because 59% of sales already run through it and every agent eventually hands the human back to your cart.
• Start a DPP data inventory if you sell batteries, textiles, or electronics into the EU, well ahead of your category deadline.
• Treat verified reviews and user-generated content as agent ranking fuel, and make them easy to crawl.

Ecommerce 2026 rewards the same discipline it always has, applied to a new reader. The numbers, from $262 billion in AI-influenced holiday sales to a projected $6.8 trillion market, all point one direction: the brands that make themselves legible, reliable, and fast for both humans and the agents shopping on their behalf are the ones that will keep the sale. A focused ecommerce growth strategy starts with your product data, not your ad budget.

Theo Werner

Paid Media Director

Theo manages Skitrate's paid-acquisition practice — Google Ads, Meta, LinkedIn, Reddit, TikTok and the smaller channels worth testing. Manages $20M+ in active client spend. Spent six years inside Google's Performance Solutions team before going agency-side. Specializes in Performance Max diagnostics, full-funnel attribution and creative testing.

• Manages $20M+ in annual client ad spend across Google, Meta, LinkedIn, Reddit, TikTok
• Ex-Google Performance Solutions specialist (Search and Performance Max)
• Designed Skitrate's full-funnel attribution methodology and creative-iteration framework
• Frequent speaker at HeroConf, SMX and Reddit's Advertising Summit

More posts by Theo Werner →