Agentic commerce statistics measure how far AI shopping agents, software that finds, compares and buys products for a person, have reached into online retail: forecasts of the market’s size, measured adoption by shoppers and platforms, and how often an agent completes a purchase. Each figure here names its source and date and says whether it is a forecast or a measurement, beside what Ottom measures on real stores.
Last updated: 13 October 2026. Every third-party figure was checked against its primary source on that date.
Market size forecasts
Projections of a future year by analyst firms. Nobody has measured these numbers, and each firm defines agentic commerce differently, which is most of why they disagree.
Forecast · $1 trillion
McKinsey forecast in October 2025 that by 2030 the US B2C retail market alone could see up to $1 trillion in orchestrated revenue from agentic commerce, with global projections reaching as high as $3 trillion to $5 trillion. #
An upper bound for 2030, not a measurement of today. Revenue an agent orchestrates is not the same thing as a checkout an agent completed.
Forecast · 15% to 25%
Bain & Company estimated in December 2025 that the US agentic commerce market could reach $300 billion to $500 billion by 2030, roughly 15% to 25% of overall e-commerce. #
Bain counts purchases initiated, influenced or completed by AI agents, and excludes journeys that only use AI for search or discovery. That is wider than an agent completing the checkout itself.
Forecast · 90%
Gartner predicted in October 2025 that by 2028, 90% of B2B buying will be AI agent intermediated, pushing over $15 trillion of B2B spend through AI agent exchanges. #
Business purchasing, not consumer retail, and a strategic prediction rather than a model of measured spend.
Adoption, measured
Things that were counted or surveyed, by the organisation named. Read who did the counting: several are a platform reporting on its own traffic.
Measured · 693.4%
Adobe Analytics measured that traffic to US retail sites from generative AI tools rose 693.4% year over year across the 2025 holiday season, 1 November to 31 December, in figures published on 7 January 2026. #
Counts shoppers clicking a link from an AI tool through to a retail site, not purchases made by an agent. Adobe notes the base of users remains modest, so a large percentage starts from a small number.
Measured · $262 billion
Salesforce estimated that AI and agents influenced $262 billion of online holiday sales worldwide between 1 November and 31 December 2025, in figures published on 8 January 2026. #
Influence, not completion. Salesforce attributes sales to AI through recommendations and engagement, extrapolated from activity on its own platform with stated market assumptions. It is not a count of purchases an agent made.
Measured · 13x
On Shopify’s Q1 2026 earnings call on 5 May 2026, president Harley Finkelstein said AI-driven traffic to Shopify stores had grown 8x year over year, and that orders from AI-powered searches had increased nearly 13x. #
One platform reporting on itself. It is the best sourced trajectory figure available and it is not an industry average.
Measured · 3x
On Shopify’s Q2 2026 earnings call on 5 August 2026, Harley Finkelstein said both AI-driven traffic and orders to Shopify stores tripled year over year in the second quarter. #
The same platform reporting on itself, against a base it has not published. A lower multiple than the quarter before is still growth, and it is not an industry average.
Measured · 22%
A Semrush survey of 1,030 US shoppers who had tried AI tools, fielded in December 2025 and published on 5 March 2026, found that 22% had completed a purchase directly inside an AI tool. #
The population is people who already use AI tools, by design. Read it as evidence that buying inside a chat is real, not as a share of all retail.
How often agents complete a purchase
Academic benchmarks of AI agents on test websites built for the purpose. They are the only published success rates for agents attempting to shop, and they are not live stores.
Research · 29%
In WebShop, a Princeton benchmark of a simulated online store published in July 2022, the best agent achieved a task success rate of 29%, against 59% for human experts. #
A research environment tested with 2022 models. It shows how hard finding and buying the right product is for software, not how often agents fail on live stores today.
Research · 14.41%
In WebArena, a Carnegie Mellon benchmark of realistic websites including an online store, published in July 2023, the best GPT-4 based agent achieved an end to end task success rate of 14.41%, against 78.24% for humans. #
The tasks span four kinds of website, not only shopping, and models have improved since. A measure of difficulty, not a failure rate for stores.
What Ottom measures on real stores
Every figure above is a forecast, a platform reporting on its own traffic, or a laboratory benchmark. None records what happens when an AI agent tries to buy from a live store. Ottom measures that: real AI agents attempt to discover, evaluate, compare and buy on each store scanned, and the aggregate is published on the benchmark pages, refreshed nightly. The sentences below are read from the same published snapshot, so they change when it does.
Across 325 stores Ottom measured at every stage in the 90 days to 13 October 2026, in every category and including stores whose category could not be determined, the median agentic readiness score was 97 out of 100. 85% of them let an AI shopping agent reach the payment step.
Of the 325 stores Ottom measured at every stage, 83% scored 80 or better at the discover stage, against 90% at the compare stage.
The most common stall across all stores is products have no GTIN or MPN to match against, found on 64% of the 345 stores Ottom scanned in the 90 days to 13 October 2026.
Apparel: Across 60 apparel stores Ottom measured at every stage in the 90 days to 13 October 2026, the median agentic readiness score was 97 out of 100. 93% of them let an AI shopping agent reach the payment step.
Beauty and personal care: Across 29 beauty and personal care stores Ottom measured at every stage in the 90 days to 13 October 2026, the median agentic readiness score was 97 out of 100. 90% of them let an AI shopping agent reach the payment step.
Electronics: Across 24 electronics stores Ottom measured at every stage in the 90 days to 13 October 2026, the median agentic readiness score was 97 out of 100. 83% of them let an AI shopping agent reach the payment step.
Food and drink: Across 31 food and drink stores Ottom measured at every stage in the 90 days to 13 October 2026, the median agentic readiness score was 97 out of 100. 81% of them let an AI shopping agent reach the payment step.
Health and wellness: Across 27 health and wellness stores Ottom measured at every stage in the 90 days to 13 October 2026, the median agentic readiness score was 97 out of 100. 78% of them let an AI shopping agent reach the payment step.
Home and furniture: Across 30 home and furniture stores Ottom measured at every stage in the 90 days to 13 October 2026, the median agentic readiness score was 97 out of 100. 93% of them let an AI shopping agent reach the payment step.
Jewellery and accessories: Across 34 jewellery and accessories stores Ottom measured at every stage in the 90 days to 13 October 2026, the median agentic readiness score was 97 out of 100. 94% of them let an AI shopping agent reach the payment step.
Outdoor and sports: Across 34 outdoor and sports stores Ottom measured at every stage in the 90 days to 13 October 2026, the median agentic readiness score was 97 out of 100. 74% of them let an AI shopping agent reach the payment step.
A store enters Ottom’s survey only if a catalogue of at least ten products can be read from it in one request, which in practice means Shopify and WooCommerce storefronts. These figures therefore describe the storefronts Ottom can read in bulk rather than online retail as a whole. The stores are a convenience sample, not a random or representative sample of online retail: a store is counted because someone scanned it or because its catalogue could be read in bulk, so these figures describe what Ottom has seen and nothing wider. The methodology sets out what is counted and what is left out.
Figures left out
Some numbers are quoted widely and could not be traced to a page the organisation itself published. They are not on this page, because a figure whose source cannot be checked is the kind this subject has too many of.
Morgan Stanley’s widely quoted forecast for US agentic commerce in 2030, and the AlphaWise survey of AI-assisted purchases reported alongside it. We found both only in press coverage, with no Morgan Stanley page carrying the figures, so neither number is repeated here.
Nobody has measured it yet, so every market size figure is a forecast. McKinsey forecast in October 2025 that by 2030 the US B2C retail market alone could see up to $1 trillion in orchestrated revenue from agentic commerce, with global projections reaching as high as $3 trillion to $5 trillion. Bain & Company estimated in December 2025 that the US agentic commerce market could reach $300 billion to $500 billion by 2030, roughly 15% to 25% of overall e-commerce. Most of the gap between the two is definition: which purchases count as an agent’s.
How many shoppers buy through AI agents today?
A Semrush survey of 1,030 US shoppers who had tried AI tools, fielded in December 2025 and published on 5 March 2026, found that 22% had completed a purchase directly inside an AI tool. Most other adoption figures count traffic or influence rather than purchases an agent completed. Adobe Analytics measured that traffic to US retail sites from generative AI tools rose 693.4% year over year across the 2025 holiday season, 1 November to 31 December, in figures published on 7 January 2026.
What is the success rate of AI agents at checkout?
There is no published industry figure for live stores. The success rates that exist come from research benchmarks on test websites. In WebShop, a Princeton benchmark of a simulated online store published in July 2022, the best agent achieved a task success rate of 29%, against 59% for human experts. In WebArena, a Carnegie Mellon benchmark of realistic websites including an online store, published in July 2023, the best GPT-4 based agent achieved an end to end task success rate of 14.41%, against 78.24% for humans. Ottom publishes how far agents get on real stores, including the share where an agent reached the payment step, on its benchmark pages once enough stores have been measured.
Which of these numbers are forecasts and which are measurements?
Every market size figure for 2028 or 2030 is a forecast. Traffic, survey and earnings call figures are measurements, made by the organisation named, and several are a platform reporting on itself. Each figure on this page is labelled Forecast, Measured or Research beside the number.