Agentic Commerce Market Size and Trends 2026: Every Forecast to 2030, Compared
Agentic commerce is still small in 2026 and forecast to be very large by 2030. EMARKETER expects about 1.5% of US retail ecommerce, or $20.57 billion, to run through AI platforms this year. By 2030, the major forecasts range from $190 billion to $1 trillion in the US and from $1.5 trillion to $5 trillion globally. The spread is wide because each firm measures something different. This report compares every major forecast, explains what each one counts, and sets out the trends that will decide which numbers turn out right.
Updated 23 October 2026. We re-checked every figure in this report against its original source. We removed numbers we could not trace, added the 2026 forecasts from Juniper Research, Edgar, Dunn & Co, Gartner and Mastercard, and corrected our description of ChatGPT checkout. Every forecast below links to the firm that published it.
Agentic Commerce Market Size at a Glance (October 2026)
| Question | Short answer | Source |
|---|---|---|
| How much US retail ecommerce runs through AI platforms in 2026? | $20.57 billion, about 1.5% | EMARKETER |
| How big is the agentic commerce technology market? | $5.7 billion in 2025, $7.7 billion in 2026, $65.5 billion by 2033 | Grand View Research |
| US agentic commerce by 2030 | $190 billion to $1 trillion | Morgan Stanley, Bain, McKinsey |
| Global agentic commerce by 2030 | $1.5 trillion to $5 trillion | Juniper Research, Edgar, Dunn & Co, McKinsey |
| Share of ecommerce by 2030 | About 10% to 29% | Morgan Stanley, Bain, Edgar, Dunn & Co |
| Shoppers willing to let AI make the purchase decision | Up to 11% | Gartner |
Agentic Commerce Market Forecasts to 2030, Compared
| Firm (published) | Forecast | Geography | What it counts |
|---|---|---|---|
| McKinsey (Oct 2025) | $900 billion to $1 trillion by 2030; $3–5 trillion globally | US B2C retail; global | Revenue that agents “orchestrate”. Goods only: no services, no B2B |
| Bain & Company (Dec 2025) | $300–500 billion, 15–25% of ecommerce, by 2030 | US | Purchases initiated, influenced or completed by third-party or retailer-hosted agents. Excludes journeys that only use AI search |
| Morgan Stanley (Dec 2025) | $190–385 billion, 10–20% of ecommerce, by 2030 | US | Spending by “agentic shoppers”. 10% is the likely case, 20% the optimistic one |
| Juniper Research (Apr 2026) | $1.5 trillion in 2030 | Global | Agentic commerce spend, growing from pilot deployments in 2025 and 2026 |
| Edgar, Dunn & Co (Apr 2026) | $2.9 trillion by 2030, 29% of ecommerce; 185% a year from 2026 | Global | Retail transactions completed with an AI agent or an LLM |
| EMARKETER (Dec 2025 forecast) | $20.57 billion, 1.5% of retail ecommerce, in 2026 | US | Retail sales through AI platforms. A current-year figure, not a 2030 one |
| Mastercard (Sep 2026) | More than 300 million online shoppers routinely using AI agents to shop and pay by 2030 | Global | People, not dollars. Based on futurists’ predictions and an Opinium survey |
| Grand View Research | $5.7 billion (2025) to $65.5 billion (2033), 35.7% a year | Global | Revenue of agentic commerce technology vendors, not sales through agents |
Why the Forecasts Differ by a Factor of Five
Three choices drive almost all of the spread:
- What counts as “agentic”. McKinsey’s “orchestrated” revenue and Bain’s “initiated, influenced or completed” both include purchases where an agent did the research and a person clicked buy. Edgar, Dunn & Co counts transactions completed with an agent. A forecast that only counts purchases an agent executes on its own would be much smaller.
- Where. Three of the forecasts are US-only. McKinsey’s US figure is between a fifth and a third of its global range.
- What is excluded. McKinsey counts goods only and leaves out services and B2B. Gartner forecasts separately that AI agents will intermediate more than $15 trillion of B2B purchases by 2028, which none of the retail figures include.
Use the range as the answer, and quote the definition with any single number. “$300–500 billion of US purchases that agents initiate, influence or complete” means something different from “$1.5 trillion of agentic commerce spend”.
Market-Research Reports vs Transaction Forecasts
Reports that size “the agentic commerce market” in single-digit billions, such as Grand View Research’s $5.7 billion for 2025, measure revenue for the companies selling agentic commerce software and services. The consulting forecasts above measure the value of goods bought through agents. Both are legitimate. They answer different questions, and mixing them produces claims like “a $5 billion market worth $5 trillion”.
The same applies to the broader agentic AI market. IDC forecasts that agentic AI will exceed 26% of worldwide IT spending, about $1.3 trillion, in 2029, with AI spending growing 31.9% a year between 2025 and 2029. That covers all enterprise uses of agents, not shopping.
What the 2026 Data Shows So Far
The forecasts describe 2030. The measured data describes a market where AI shapes a lot of discovery and completes few purchases.
| Measure | Finding | What it counts |
|---|---|---|
| Orders influenced by AI during Cyber Week 2025 | 20% of global orders; $67 billion of $336.6 billion in sales (Salesforce, Dec 2025) | Orders influenced by AI recommendations and service agents, not purchases agents made |
| AI-referred traffic to US retail sites, holiday 2025 | Up 693.4% year on year (Adobe, Jan 2026) | Clicks from AI services to retail sites, from a small base |
| AI traffic and orders on Shopify, Q1 2026 | Traffic up 8 times year on year; orders from AI search up nearly 13 times (Shopify) | Shopify merchants only. In Q3 2025 the figures were 7 times and 11 times since January (TechCrunch) |
| Share of US retail ecommerce through AI platforms, 2026 | About 1.5%, $20.57 billion (EMARKETER) | Forecast for the full year |
| Consumers who expect to use AI shopping agents within 12 months | 60% (Kearney, 2025) | Intent, from two US surveys of about 750 consumers each |
| Consumers willing to let AI make purchase decisions | Up to 11%; 31% would let AI narrow choices for household supplies (Gartner, Jan 2026 survey of 322) | Willingness, in lower-stakes categories |
| AI users who bought entirely through an agent | 8% (L.E.K., Apr 2026 survey of 2,650) | Purchases made without visiting a retailer’s site |
AI traffic has also become more valuable. When Adobe reported AI traffic to US retail sites up 1,200% in February 2025, the comparison was with July 2024, seven months earlier, and AI visitors were 9% less likely to convert than other traffic. By August 2026, Adobe found AI-referred visitors converting 48% better than non-AI traffic. The visits got better before they got numerous: L.E.K. still puts AI referrals at under 0.5% of inbound traffic.
For the 2026 holiday season, including Adobe’s forecast of 130% more AI-referred traffic and which assistants can check out, see our agentic commerce Q4 2026 tracker. We will add Adobe’s and Salesforce’s holiday results there after Cyber Monday.
Agentic Commerce Trends in 2026
1. Checkout Moved Into Some Assistants and Out of Others
Four assistants can complete a purchase in the US as of October 2026: Google AI Mode and Gemini (through the Universal Commerce Protocol, now with a multi-retailer Universal Cart), Microsoft Copilot (Copilot Checkout, since January 2026), Meta Muse (the agent buys after the shopper approves the total) and Perplexity (Instant Buy with PayPal).
ChatGPT went the other way. OpenAI launched Instant Checkout in September 2025 for US Etsy sellers and planned Shopify merchants next. On 24 March 2026 it let merchants use their own checkout instead and focused ChatGPT on product discovery. Shopify now sends merchant catalogs to ChatGPT, Copilot, AI Mode and Gemini through Agentic Storefronts, switched on by default. The Q4 2026 tracker has the full platform table.
2. Protocols Became the Plumbing
Two open protocols carry most agent-to-merchant traffic. OpenAI and Stripe’s Agentic Commerce Protocol (ACP) now covers product discovery as well as checkout, and Microsoft cites it for Copilot Checkout. Google announced the Universal Commerce Protocol (UCP) at NRF in January 2026, and it runs checkout in AI Mode and Gemini. On the payment side, Visa, Mastercard and Stripe each have ways for agents to pay with a token instead of a card number. Our guide to the agentic commerce stack covers ACP, UCP, AP2 and agent identity in detail.
For a merchant, the practical effect is that one clean product feed can serve several assistants, much as one payment processor serves every card network.
3. Retailers Split on Outside Agents
Amazon blocks outside shopping agents, including Meta Muse since September 2026, and promotes its own assistant, Alexa for Shopping. Walmart took the opposite approach in October 2026: its Agent Gateway lets verified outside agents buy within spending limits, without ever receiving the customer’s card details. Which approach wins will shape how much of the forecast volume flows through independent assistants and how much stays inside retailers’ own agents.
4. Discovery Matters More Than Checkout, For Now
Most shoppers still want to finish the purchase themselves. In Attentive’s August 2026 survey, 60% preferred to complete the purchase on the seller’s site or app, against 17% who preferred the chatbot, and 93% of AI-assisted shoppers looked for proof outside the AI before buying. Meta’s Muse, Google and the rest are building checkout, but in 2026 the commercial value of agents lies mostly in which products they recommend. Our 2030 predictions for Meta Muse look at how that could change.
5. Product Data Decides Who Agents Can Recommend
If an agent cannot read a product’s price, stock, specifications and return policy, it recommends something else. That is algorithmic invisibility: the shopper never sees the product, so there is no click to lose. Merchant readiness lags. PayPal’s April 2026 merchant survey found only about one in five merchants had 80% or more of their catalog available as structured, machine-readable data. Our guide to optimizing ecommerce product pages for AI covers the fields agents use.
6. China Is Further Along
In China, Alibaba connected its Qwen assistant to Taobao, Alipay, Fliggy and Amap, and Qwen handled more than 120 million orders in six days during the 2026 Lunar New Year. That is the closest thing yet to the forecasts’ 2030 world, in one ecosystem. We cover it in China’s AI internet and agentic marketing.
Consumer Segments and Adoption Barriers
Kearney’s report Agentic commerce: from brand loyalty to bot logic (August 2025) divides US consumers into four groups by how they approach AI shopping agents:
| Segment | Share | What they want | Likely adoption |
|---|---|---|---|
| Tech-forward early adopters | ~15% | To automate routine purchases and save time | Fastest |
| Price-sensitive pragmatists | ~35% | Price comparisons, discounts and savings guarantees | Fast, if the savings are visible |
| Privacy-conscious skeptics | ~30% | Control over decisions and data | Slow, needs trust |
| Routine-loyal traditionalists | ~20% | Familiar brands and human interaction | Slowest |
Trust is the barrier every source names. Juniper Research calls it the number one obstacle to agentic commerce. Gartner’s survey found willingness to let AI make purchase decisions topped out at 11%, even in lower-stakes categories such as personal care and household supplies, while far more people would let AI narrow the options. The near-term opportunity is agents that research and compare; agents that buy on their own will grow as payment controls, spending limits and returns become easier to trust.
Market Constraints and Risks
Several things could keep the market at the low end of the forecasts:
- Platform decisions. OpenAI’s checkout reversal and Amazon’s agent blocking show that a single company can change the channel in a quarter.
- Trust and fraud. Shoppers want visible control, and merchants need to verify which agents are legitimate before letting them pay.
- Regulation. Rules on automated purchasing, liability and data protection, including the EU AI Act, are still settling.
- Data quality. Retailers without clean, structured product data will be invisible to agents, which concentrates the gains among larger, better-organized sellers.
Our agentic commerce guide turns these into a risk table with a metric and a trigger for each.
What the Forecasts Mean for Brands
Whichever forecast proves right, the purchase decision is moving into AI answers before it reaches a checkout. Three things follow:
- Measure whether agents recommend you. Track the prompts your customers use in ChatGPT, Gemini, Perplexity, Claude and Google AI Mode, and how often each one names your products. Our guide on how to measure AI visibility explains the metrics.
- Make product data complete and consistent. Price, stock, specifications, shipping and returns should match across your feed, your site and Google Merchant Center.
- Earn independent evidence. Agents weigh reviews, comparisons and forum discussion. Brands with more of it get recommended more often.
Sanbi tracks brand and product visibility daily across ChatGPT, Gemini, Perplexity, Claude and Google AI Mode. You can run a free AI visibility audit to see where you stand today.
First published: 2 March 2026. Updated: 23 October 2026.
Author: Aditya Sinha, Founder, Sanbi.ai
Sources
- Forecasts: McKinsey, The agentic commerce opportunity (Oct 2025; coverage by Digital Commerce 360); Bain, 2030 forecast for US retail (Dec 2025; coverage by Digital Commerce 360); Morgan Stanley, Here come the shopping bots (Dec 2025); Juniper Research (Apr 2026); Edgar, Dunn & Co (Apr 2026); EMARKETER; Mastercard (Sep 2026); Grand View Research; IDC (Aug 2025)
- Measured data and surveys: Salesforce Cyber Week 2025; Adobe holiday 2025 results; Adobe, February 2025 AI traffic; Shopify, how agentic commerce works; TechCrunch on Shopify Q3 2025; Gartner consumer survey (May 2026); Kearney (Aug 2025; coverage by Digital Commerce 360); L.E.K.; Attentive; PayPal merchant pulse
- Platforms: see the sources in our Q4 2026 checkout tracker
Frequently Asked Questions
It depends on what you count. Measured as sales, EMARKETER forecasts $20.57 billion of US retail ecommerce, about 1.5%, will run through AI platforms in 2026. Measured as revenue for the vendors selling agentic commerce technology, Grand View Research puts the global market at $7.7 billion in 2026, up from $5.7 billion in 2025. Juniper Research describes 2025 and 2026 as a period of pilot deployments.
Forecasts for 2030 range from $190 billion to $1 trillion in the US and from $1.5 trillion to $5 trillion globally. Morgan Stanley expects $190–385 billion of US ecommerce spending, Bain $300–500 billion, and McKinsey $900 billion to $1 trillion of US retail revenue orchestrated by agents. Globally, Juniper Research forecasts $1.5 trillion, Edgar, Dunn & Co $2.9 trillion and McKinsey $3–5 trillion.
Between about 10% and 29%, depending on the forecast. Morgan Stanley puts agentic shoppers at 10% of US ecommerce in its likely case and 20% in its optimistic case. Bain estimates 15–25% of US ecommerce. Edgar, Dunn & Co forecasts 29% of global ecommerce. For comparison, EMARKETER puts the 2026 share of US retail ecommerce through AI platforms at about 1.5%.
Each firm counts something different. McKinsey sizes revenue that agents orchestrate, for goods only. Bain counts purchases that agents initiate, influence or complete. Morgan Stanley sizes spending by agentic shoppers. Edgar, Dunn & Co counts transactions completed with an AI agent. Market-research firms such as Grand View Research measure vendor revenue, which is why their figures are in billions rather than trillions.
McKinsey estimates that by 2030 agents could orchestrate $900 billion to $1 trillion of US B2C retail revenue, and $3 trillion to $5 trillion globally. The figures cover goods only and exclude services and B2B. McKinsey published the estimate in October 2025, based on US Census data and forecast adoption rates.
Five stand out. Checkout moved into some assistants (Google AI Mode and Gemini, Microsoft Copilot, Meta Muse and Perplexity), while ChatGPT moved purchases back to merchants' own checkouts in March 2026. Protocols such as ACP and UCP became the plumbing. Retailers split on outside agents, with Amazon blocking them and Walmart opening an Agent Gateway. Discovery still matters more than checkout. Product data quality decides which brands agents can recommend.
North America leads Western markets: Grand View Research gives it 38.2% of global agentic commerce technology revenue in 2025, and McKinsey expects the US to account for $900 billion to $1 trillion of its $3–5 trillion global estimate. China is further ahead in practice. Alibaba's Qwen assistant handled more than 120 million orders in six days during the 2026 Lunar New Year.
Agentic shopping is when an AI assistant researches, compares and, with the shopper's permission, buys products on the shopper's behalf. The shopper describes a need, the agent searches product data, reviews and prices, presents options and can complete checkout. As of October 2026, Google AI Mode and Gemini, Microsoft Copilot, Meta Muse and Perplexity can complete a purchase in the US; ChatGPT hands the purchase to the merchant's site.
Algorithmic invisibility is when an AI shopping agent never proposes a brand's products because it cannot read or trust the brand's data. The shopper never sees the product, so there is no click to lose. Common causes are incomplete product feeds, missing structured data, inconsistent prices or stock across channels, and too few independent reviews and comparisons.