Agentic Commerce: 2026 Statistics & Benchmarks You Need to Know

Refreshed and updated for 2026: All statistics are direct first-party sources where available. The market-size estimates measure different things, and shouldn’t be read as one figure. McKinsey counts orchestrated revenue across the whole commerce chain, eMarketer counts only checkout that happens inside an AI platform, and Gartner counts B2B procurement.

AI shopping agents now discover consumer products, build baskets, and complete some purchases on their own. The figures below show where consumer trust, agent reliability, merchant readiness, and the law each stand right now.

Agentic AI market size and growth forecasts and statistics

There’s wide disagreement on how big the agentic commerce era will get, and most of it traces to definition differences. But almost all agree that conversational commerce will see significant AI investment over the next few years.

• McKinsey puts the US B2C opportunity at $900 billion to $1 trillion by 2030, with a global range of $3 trillion to $5 trillion. Those figures cover goods only, and exclude services and the B2B market.
• Morgan Stanley takes a stricter line that requires meaningful autonomous action, and arrives at $190 billion to $385 billion of US ecommerce by 2030, or 10% to 20% of the online total. 
• Bain, which includes agent-influenced purchases, projects $300 billion to $500 billion for the US, or 15% to 25% of ecommerce.
• eMarketer measures the narrowest slice, counting only checkout inside an AI platform, and forecasts $20.57 billion in US retail spending for 2026, around 1.5% of total ecommerce.

Then there’s B2B, which dwarfs the consumer side. 

• Gartner expects 90% of B2B buying to run through AI agents by 2028, pushing over $15 trillion through agent exchanges. 
• Gartner also projects that 20% of monetary transactions will be programmable by 2030, giving agents economic agency.

Statistics on AI traffic to retail sites

Shoppers are reaching retail sites through AI tools, and they convert better once they get there.

• During the 2025 holiday season, AI-driven traffic to US retail sites rose 693.4% year over year, according to Adobe. 
• The growth held into 2026: AI traffic was up 393% year over year in the first quarter, peaking at 1,151% in December. 
• Measured from October 2024, when Adobe started tracking it, AI-referred traffic has grown 1,324%, and in May 2026 it was still up 138% year over year.

The conversion story reversed inside a year. 

• In March 2025, AI traffic converted 38% worse than non-AI traffic. By March 2026, it converted 42% better, a record in Adobe’s data. 
• AI-referred shoppers also spent 48% longer on site, browsed 13% more pages, and generated 37% more revenue per visit. 
• By May 2026, those shoppers spent 53% longer on site and browsed 23% more pages.

Salesforce found a similar pattern in the run-up to the holidays:

• AI assistant traffic converted 700% better than social media referrals and 200% better than traditional search.

But a lot of retail content still can’t be read by a machine:

• Adobe scored the average US product page at 66% machine-readable, which means around 34% of the content on the page where people decide to buy can’t be parsed by an agent.

Statistics on consumer adoption and trust in agentic commerce

Adoption is climbing fast, and trust is the brake. Today’s shopper is willing to hand over routine baskets first, but they want hard limits on agentic capability before they do.

• In the UK, the share of shoppers using AI assistants doubled from 12% to 28% in a year, per Adyen, and 44% would let an agent handle the whole process once they’ve set a budget and brand. 
• In the US, Adyen found usage more than doubled from 12% to 35%, and more than half (51%) of shoppers would let AI handle the entire process including the final purchase. 
• Among US AI shoppers, 66% say it saves time and 62% say it cuts through online noise.
• Adobe’s survey put usage at 39% of US consumers having shopped with AI, of whom 85% said it improved their experience, and 66% said they believe AI tools provide accurate results. 
• In Europe, a McKinsey survey across France, Germany, and the UK found 84% of consumers use AI tools in everyday life, with 63% using AI to compare brands, prices, and reviews, and 38% using it for product research.

Delegation starts with the boring stuff:

• Checkout.com found 47% of consumers would let an agent handle repetitive purchases, rising to two-thirds among 25-to-44-year-olds.
• A separate study put UK comfort with routine shopping at 40%. 
• Brand loyalty bends, too: 57% would let an agent switch brands for better value. 
• On average, consumers would let an agent spend £177 per purchase before asking again.
• And their top conditions for delegating are spending caps (30%), instant revocation (29%), and easy cancellation (28%).
• 75% of merchants say giving customers the real-time ability to revoke permissions will be critical to agentic commerce adoption.

The hesitation runs deep:

• A third of consumers (33%) expect at least 10% of purchases to be AI-driven within a year, yet 24% say they’ll never delegate purchases to AI and 27% trust no organisation to operate an agent. 
• Earlier Checkout.com data found 42% worry about losing control of what an agent buys and 28% cite a lack of transparency. 
• There’s a clear lean toward retailer-run agents: a Bain survey found consumers trust retail-owned agents three times more than third-party agents to complete a transaction.

Some of this delegation predates the current wave:

• Around 23% of US Amazon shoppers had at least one active Subscribe & Save order in 2024, an early proxy for the kind of standing automated purchase agents now promise. 
• And McKinsey found 44% of AI-search users already call it their primary and preferred way to search the internet, against 31% who prefer traditional search.

AI shopping agent performance and reliability statistics

Demos run fine, but daily use doesn’t. The benchmarks show agents handling bounded tasks well and stumbling when a purchase needs judgement.

• On a careful shopping benchmark, the strongest model scored 17.76% against human experts’ 30.02%, and passed safety checks only 35.42% of the time. 
• On WebMall, a four-shop comparison test, the strongest agent handled add-to-cart and checkout without trouble but completed under 65% of harder jobs like finding the cheapest offer across shops. 
• On DeepShop, the top system reached only 20% on hard queries.

Reliability also breaks down under repetition:

• Research estimates enterprises generally need agent failure rates below 5%.
• Yet GPT-4-based agents that succeed about 60% of the time on a single attempt drop to roughly 25% across eight consecutive runs.

The capability curve points up, though:

• The length of task an agent can finish at even odds has been doubling roughly every 7 months since 2019, according to METR. 
• And Gartner expects AI agents to outnumber human sellers 10-to-1 by 2028, while fewer than 40% of sellers report that agents improved their productivity.

Statistics on holiday commerce and artificial intelligence agents

The 2025 holidays gave the clearest read yet on agents in live digital commerce, drawn from Salesforce’s analysis of more than 1.5 billion shoppers and surveys from platforms like Checkout.com.

• During the 2025 holiday season, AI and agents drove $262 billion in global online sales. 
• Forecasts for 2030 range from $190 billion (Morgan Stanley) to $5 trillion (McKinsey), depending on what each firm counts. 
• Over Cyber Week (25 November to 1 December), global sales hit $336.6 billion, and AI and agents drove $67 billion of that, influencing 20% of all purchases. 
• Across the full season (November to December), global online sales reached $1.29 trillion.
• Per Checkout.com, 47% of consumers planned to use an AI agent for their Christmas shopping. 
• A third said they planned to rely on agentic tools over Black Friday.

Brands that ran their own agentic AI systems pulled ahead:

• Companies deploying an AI agent(s), including Pandora, SharkNinja, and Funko, saw a 59% higher growth rate, averaging 6.2% year over year against 3.9%. 
• Shoppers referred from AI search channels converted nine times more often than those arriving via social media.

Customer service shifted as much as sales:

• December saw a 66% jump in agentic AI service conversations over November, and agents handled 142% more tasks, such as updating addresses and starting returns, than in the prior two months. 
• Salesforce reported powering 61 million orders on its commerce platform with 100% uptime through the week.

Agentic commerce: Merchant and payment infrastructure readiness statistics

The rails are arriving faster than stores can lay track to meet them. Most merchants know agents are coming, and few are built for it.

• The shared standards are scaling. Anthropic’s Model Context Protocol, which lets agents connect to tools and data, has roughly 10,000 public servers and SDKs that see 97 million downloads a month. 
• Google’s Agent Payments Protocol launched with 60+ partners. 
• Around 1 million Shopify merchants are anticipated to sell through ChatGPT via the Agentic Commerce Protocol.

P.S. ACP isn’t the only protocol that powers AI assisted shopping. UCP, the Universal Commerce Protocol (by Google), also powers AI shopping by autonomous agents. More on payment infrastructure and agentic systems for product discovery in Part 2 of my agentic commerce series.

Early platform data shows demand:

• Shopify reported that AI-driven traffic to its stores grew 8x year over year in the first quarter of 2026
• Orders from AI-powered searches were up nearly 13x.
• Orders attributed to AI-powered search carried 14% higher average order values (AOV) compared to organic search.
• Traffic from Catalog-powered AI searches converted 2x more than traffic from a general AI search.

But readiness lags:

• In the UK, only 15% of the top 100 retailers say their payment systems are prepared for agent transactions, even though 49% are investing in agentic AI. 
• A Payments Association survey found 58% of UK online merchants believe agents have already reached their platforms, yet only 3% of transactions involve one today. 
• Checkout.com put merchant preparation higher, with 89% of merchants actively getting ready and 72% agreeing consumers will adopt faster than merchants can.

The friction has turned litigious:

• Amazon said it spent more than $5,000 and many engineering hours blocking Perplexity’s agent, then changed its seller terms to require every AI agent to identify itself. 
• Its own Buy for Me feature is expected to drive over $10 billion in annual sales. 
• That feature drew a backlash when more than 500,000 products were scraped from outside brand sites and listed without consent, affecting over 180 sellers.

Statistics on AI agent influence on purchasing decisions

Sellers don’t need to persuade you any more. They can persuade the agent, and the tactics that work on a model differ from the ones built for human agents.

• In a Princeton experiment of 2,012 people, a conversational agent pushed people to select sponsored products 61.2% of the time, up from 22.4% under plain search. 
• When the model hid its intent, people spotted the steering less than 10% of the time.

The levers are different from human marketing:

• Harvard Business Review research found that countdown timers, scarcity, and strike-through prices don’t reliably move AI agents and can even lower selection.
• Meanwhile star ratings and a lower price influence agents more consistently. 
• Separate work shows agents favour items placed higher in a list, and that a product description rewritten to exploit known model biases can swing the pick.

For all that, the seller-side modelling is still early:

• Gartner places the customer digital twin at the earliest stage of its hype cycle, which fits a market where today’s agents mostly ask rather than decide.

Statistics on agentic commerce authority, fraud, and security

Agents add new surfaces for impersonation and fraud, on top of a base of card fraud that’s already large.

• In the UK, remote purchase fraud losses reached £423.5 million in 2025, per UK Finance. 
• In the US, the FTC reported people lost about $16 billion to fraud in 2025, a 25% year-over-year increase.

There are precedents for binding authority to time and scope:

• UK open banking rules from the FCA require third-party providers to refresh customer consent at least every 90 days when accessing account information. 
• And the FIDO Alliance has begun work on standards for trusted AI agent interactions, including agent authentication and payments working groups.

Statistics on agentic commerce liability and regulation

The recovery system runs on rules written for humans tapping cards, and the gaps are starting to show.

• US generative-AI lawsuits climbed 978% between 2021 and 2025, per Gallagher Re. 
• In the UK, the CMA told businesses the same consumer law applies whether a customer deals with a human or an agent, and that breaches under the Digital Markets, Competition and Consumers Act can bring fines of up to 10% of worldwide turnover.

The US has moved the same way:

• California chaptered AB 316 in October 2025, barring defendants who developed or used AI from arguing the AI autonomously caused the harm. 

Read the full agentic commerce series

These stats don’t exist in a vacuum. Each one connects to a structural layer of how agentic commerce actually works, from execution and payment infrastructure to authority, liability for agent capability, and governance.

My seven-part series breaks it all down:

• Part 1: Execution — how autonomous AI agents browse, compare, and buy
• Part 2: Infrastructure — the protocols and rails underneath agent transactions
• Part 3: Preferences — what consumers will and won’t delegate, and on what terms
• Part 4: Judgement — how AI commerce agents make purchase decisions, and who can influence them
• Part 5: Authority — identity, authentication, and the question of who authorised the purchase
• Part 6: Liability — what happens when an agent gets it wrong
• Part 7: Governance — the regulatory and policy layer around AI capabilities and agent interaction

---

About the author

Mo Shehu, PhD is a practitioner and researcher of generative AI systems and agentic commerce. Earlier in his career, he managed social media and influencer marketing across eight P&G brands, including Always, Olay, Pampers, and Vaseline, with monthly budgets above $430,000 and a combined reach of over 10 million. He holds a PhD in Informatics focused on social media analytics.

He has worked on retail brands across menswear and consumer health, run paid and organic ecommerce campaigns, and published free industry reports on AI policy, agentic AI governance, and the economics of delegated decision-making on consumer behavior and customer engagement.

He is the author of the seven-part series on AI driven commerce series and the builder of Agent Kit, a WordPress plugin that adds agentic commerce signals to WooCommerce storefronts. 

→ Connect with Mo on LinkedIn