Skip to content

Guide7 min readPublished

Agentic commerce statistics 2026: what the published figures say

First-party figures from Adobe, Cloudflare, Microsoft, Salesforce, Visa and Morgan Stanley, each quoted from the publisher that measured it, with its date and what it measures. We never add one figure to another or convert it.

"While the base of users remains modest." Adobe put that sentence straight after its headline figure for the 2025 holidays, a 693.4% rise in visits from generative AI tools to US retail sites. Of the two, guess which one ends up on slides.

The published, first-party figures say that clicks from AI assistants to stores are growing fast from a small base, that those visitors now convert better than non-AI traffic by Adobe's measure, and that AI agents are making far more requests to websites than a year ago. No publisher we read prints a count of purchases made by AI agents, and the large dollar numbers are forecasts or measures of "AI" broadly, recommendation engines included.

Six publishers made the cut: Adobe, Cloudflare, Microsoft, Salesforce, Visa and Morgan Stanley. Every figure here is from its publisher's own page, with its date and what it measures. We do not add, convert or combine any of them, and anything we could only find second-hand stayed out.

Which agentic commerce statistics come from the people who measured them?

Thirteen figures from six publishers. Highlight one kind to see what each kind of figure can tell you:

First-party agentic commerce figures: who measured what, and when
Highlight

Showing all 13 figures.

  • +693.4%Adobe · 7 Jan 2026 · Clicks from AI to stores

    Measures: Visits to US retail sites from generative AI tools, “shoppers clicking on a link to a retail site”, 1 November to 31 December 2025, against the year before.

    What it does not tell you (our reading): Growth, not size. Adobe adds: “While the base of users remains modest”.

    “traffic to retail sites from generative AI tools (shoppers clicking on a link to a retail site) increased by 693.4% compared to the year prior.”

    Source: Adobe: 2025 holiday shopping season (7 Jan 2026)

  • +71%Adobe · 29 Oct 2026 · Clicks from AI to stores

    Measures: AI traffic to US retail sites in September 2026, “measured by users clicking on a link”, year over year.

    What it does not tell you (our reading): One month, US only, and again growth rather than a share of all visits.

    “In September 2026, AI traffic to U.S. retail sites (measured by users clicking on a link) grew 71% year-over-year (YoY).”

    Source: Adobe: AI traffic drives 83% higher revenue per visit (29 Oct 2026)

  • 90% better conversionAdobe · 29 Oct 2026 · Purchases and conversion

    Measures: How often a visit from an AI source became a purchase, against non-AI traffic “(such as email, paid search, and social)”, September 2026.

    What it does not tell you (our reading): A comparison of rates, not a count of sales. In August 2025, Adobe says, the same traffic “converted 9% worse”.

    “In September, AI traffic converted 90% better (visits that become purchases) than non-AI traffic, the highest level Adobe has observed to date.”

    Source: Adobe: AI traffic drives 83% higher revenue per visit (29 Oct 2026)

  • +1,700%Cloudflare · 30 Sep 2026 · Bots and agents on sites

    Measures: Daily requests from AI agents, “software acting on behalf of people”, on Cloudflare’s network, over a year.

    What it does not tell you (our reading): Requests, not shoppers or orders, and only on sites behind Cloudflare.

    “Over the past year, daily requests from AI agents on our network grew by more than 1,700%.”

    Source: Cloudflare: The Internet has a second audience (30 Sep 2026)

  • Under 1% vs 17%Cloudflare · 30 Sep 2026 · Bots and agents on sites

    Measures: Share of sites on Cloudflare that block search crawlers, against the share that block training crawlers.

    What it does not tell you (our reading): All sites on Cloudflare, not stores alone.

    “Fewer than 1% of sites on Cloudflare block search crawlers, while 17% block training.”

    Source: Cloudflare: The Internet has a second audience (30 Sep 2026)

  • 53% more purchasesMicrosoft · 8 Jan 2026 · Purchases and conversion

    Measures: Purchases within 30 minutes of a shopping journey that included Copilot, against journeys without it. Footnote: “Microsoft Internal Data, August 2025”.

    What it does not tell you (our reading): Microsoft’s own product, measured by Microsoft, on Microsoft Advertising’s blog.

    “Journeys that include Copilot led to 53% more purchases within 30 minutes of interaction compared to those without.”

    Source: Microsoft Advertising: Copilot Checkout and Brand Agents (8 Jan 2026)

  • $262 billionSalesforce · 8 Jan 2026 · Purchases and conversion

    Measures: 2025 holiday sales that “AI and agents” drove (“20% of all retail sales”) “through personalized recommendations and deeper customer engagement”, measured from the activity of “more than 1.5 billion global shoppers”.

    What it does not tell you (our reading): It counts sales influenced by personalised recommendations too, so it is not money spent by shopping agents.

    “AI and agents powered a massive portion of the holidays, driving 20% of all retail sales and fueling $262 billion in revenue through personalized recommendations and deeper customer engagement.”

    Source: Salesforce: 2025 holiday shopping data (8 Jan 2026)

  • Nine times more oftenSalesforce · 8 Jan 2026 · Purchases and conversion

    Measures: How often 2025 holiday shoppers referred from AI-powered search channels converted, against shoppers referred from social media.

    What it does not tell you (our reading): A comparison with social referrals only, and a rate rather than a count of sales.

    “shoppers referred to retailer websites from AI-powered search channels converted nine times more often than those coming through social media referrals.”

    Source: Salesforce: 2025 holiday shopping data (8 Jan 2026)

  • 39%Adobe · 17 Mar 2025 · Surveys of people

    Measures: US consumers in Adobe’s survey of 5,000 who had used generative AI for online shopping.

    What it does not tell you (our reading): A survey answer from early 2025, about using AI at all while shopping.

    “In Adobe’s survey of 5,000 U.S. consumers, 39 percent have used generative AI for online shopping, with 53 percent planning to do so this year.”

    Source: Adobe: Traffic from generative AI sources jumps 1,200 percent (17 Mar 2025)

  • 47%Visa · 18 Dec 2025 · Surveys of people

    Measures: US shoppers who use AI tools “for at least one shopping task”, from a Morning Consult survey “among a sample of 1,000 adults in 12 markets”, October 2025.

    What it does not tell you (our reading): A price comparison counts as a task. Visa calls these US shoppers, but its footnote describes a 12-market sample. It says nothing about agents paying.

    “nearly half of U.S. shoppers (47 percent) now use AI tools for at least one shopping task”

    Source: Visa: Visa and Partners Complete Secure AI Transactions (18 Dec 2025)

  • Roughly 23%Morgan Stanley · 8 Dec 2025 · Surveys of people

    Measures: Americans who “bought something via AI in the past month”, estimated from an AlphaWise survey.

    What it does not tell you (our reading): An estimate from adoption rates, not a measured count. “Via AI” is not defined on the page.

    “Based on adoption rates, Morgan Stanley estimates that roughly 23% of Americans bought something via AI in the past month”

    Source: Morgan Stanley: Here Come the Shopping Bots (8 Dec 2025)

  • $190bn to $385bn by 2030Morgan Stanley · 8 Dec 2025 · Forecasts

    Measures: Forecast US e-commerce spending by “agentic shoppers” in 2030, “a likely 10% share of online retail” and “an optimistic outlook up to 20%”.

    What it does not tell you (our reading): A forecast. Nothing has been measured yet.

    “Morgan Stanley Research estimates that agentic shoppers could reach $190 billion to $385 billion in U.S. e-commerce spending by 2030.”

    Source: Morgan Stanley: Here Come the Shopping Bots (8 Dec 2025)

  • “millions of consumers”Visa · 18 Dec 2025 · Forecasts

    Measures: Visa’s prediction for agent purchases by the 2026 holiday season.

    What it does not tell you (our reading): A prediction with no number. The same release counts “hundreds” of completed agent transactions.

    “Visa predicts that millions of consumers will use AI agents to complete purchases by the 2026 holiday season.”

    Source: Visa: Visa and Partners Complete Secure AI Transactions (18 Dec 2025)

Each figure is quoted from the publisher’s own page and read on 4 November 2026. “What it does not tell you” is our reading. We do not add, convert or combine any of them: they measure different things, in different places, over different periods.

Read the kind before the number. A rise in clicks from ChatGPT to a store, a rise in bot requests on Cloudflare's network, and a forecast of 2030 spending are three different things, and stats pages that line them up in one list usually leave that out.

What does the 693.4% measure?

Adobe defines its AI traffic as "shoppers clicking on a link to a retail site" from a generative AI tool, measured across "over 1 trillion visits to U.S. retail sites" from 1 November to 31 December 2025, against the same weeks a year before. It is a growth rate, so it says how fast the channel grew and nothing about how big it is. A rise of 693.4% from a small base is still a small number, which is what "the base of users remains modest" tells you.

Adobe's September 2026 figure is gentler. In September 2026, "AI traffic to U.S. retail sites (measured by users clicking on a link) grew 71% year-over-year". Adobe's pages give the growth and not the base, so the size of the channel stays unknown from these figures alone.

Do visitors from AI assistants buy?

By Adobe's measure, more often than other visitors now. "In September, AI traffic converted 90% better (visits that become purchases) than non-AI traffic, the highest level Adobe has observed to date." The comparison is with non-AI traffic "such as email, paid search, and social". It was not always so: "in August 2025, AI traffic converted 9% worse".

Two other first-party figures point the same way, each from a company measuring its own product or customers:

  • Microsoft: "Journeys that include Copilot led to 53% more purchases within 30 minutes of interaction compared to those without." Its footnote reads "Microsoft Internal Data, August 2025".
  • Salesforce: shoppers "referred to retailer websites from AI-powered search channels converted nine times more often than those coming through social media referrals" during the 2025 holidays.

Salesforce's headline, "$262 billion", needs its full sentence: "AI and agents powered a massive portion of the holidays, driving 20% of all retail sales and fueling $262 billion in revenue through personalized recommendations and deeper customer engagement." Personalised recommendations count, so it is not money that shopping agents spent.

How many purchases do AI agents make?

No publisher we read says. The closest is Visa, which in December 2025 wrote that "hundreds of secure, agent-initiated transactions have now been successfully completed" with its partners. Hundreds is a pilot. The same release predicts "millions of consumers will use AI agents to complete purchases by the 2026 holiday season", a prediction from a company that runs one of the payment programmes involved, covered in Visa Intelligent Commerce vs Mastercard Agent Pay.

Morgan Stanley's "roughly 23% of Americans bought something via AI in the past month" is sometimes quoted as agent purchases. The page says it is an estimate "Based on adoption rates" from a survey, and it does not define "via AI".

What do the surveys say about shoppers?

Both surveys say many shoppers use AI for something. In Adobe's survey of March 2025, "39 percent have used generative AI for online shopping". In Visa's survey of October 2025, run with Morning Consult, "nearly half of U.S. shoppers (47 percent) now use AI tools for at least one shopping task". Comparing a price counts as a task: Visa lists uses "from price comparisons to personalized recommendations". Visa calls these US shoppers, but its footnote describes a sample of "1,000 adults in 12 markets".

The two asked different questions of different samples, so 39% and 47% do not show a trend. Neither asks whether an agent paid.

What do the forecasts say?

Morgan Stanley, December 2025: "agentic shoppers could reach $190 billion to $385 billion in U.S. e-commerce spending by 2030", which it calls "a likely 10% share of online retail" with "an optimistic outlook up to 20%". It is the only spending forecast whose publisher's page we could read; McKinsey's returned an access error, so we leave it out. Adobe's page of 29 October 2026 adds a traffic projection, "with AI traffic projected to rise 159% YoY on Thanksgiving": a projection made in October for Thanksgiving 2026, not a measurement.

A forecast is an argument with a number on the end. This one is worth reading for its reasoning, and none of the figures above can confirm or refute it yet.

Why do statistics pages disagree?

Often a figure has been passed on second-hand and has lost its date and definition on the way. A few examples from our own search:

  • Visa's 2025 release cites a 4,700% rise in AI traffic to "Adobe Data Insights, August 2025". We found no Adobe page with that figure, so it is not here.
  • Shopify's growth in AI orders is reported in the press. Shopify's own results releases for the second and third quarters of 2026 contain no such figure, so it is not here either.
  • Monthly Adobe figures for early 2026 appear on many sites; we used only the ones on Adobe's pages.

Pick a question to see what the published figures can and cannot answer:

Your question, and what the published figures can and cannot answer
Your question

Showing: How many purchases do AI agents make?

What the published figures say (our reading)

No publisher we read prints a count of purchases made by agents. Visa’s December 2025 release says “hundreds of secure, agent-initiated transactions have now been successfully completed”, which tells you the scale was small then. Morgan Stanley’s 23% is people buying “via AI”, estimated, not agents buying alone.

The figures

  • “millions of consumers” (Visa, 18 Dec 2025)
  • Roughly 23% (Morgan Stanley, 8 Dec 2025)

Sources: Visa: Visa and Partners Complete Secure AI Transactions (18 Dec 2025) and Morgan Stanley: Here Come the Shopping Bots (8 Dec 2025)

How can a store measure its own numbers?

None of the figures above is about your store. The one source that is, your server or CDN logs, is also the one that sees agents that never click through. Paste a few log lines to see which requests claim to be an AI agent:

A log line classifier: which requests claim to be an AI agent

6 of 7 lines claim to be an agent on our list.

  • 2 Training crawler
  • 2 Search crawler
  • 2 Assistant, for a user
  • 1 no agent named
LineClaims to beTo check the claim
1GPTBotTraining crawlercheck the address against the vendor’s IP list (https://openai.com/gptbot.json)
2No agent on our listA person, or an agent with a browser’s User-Agent
3OAI-SearchBotSearch crawlercheck the address against the vendor’s IP list (https://openai.com/searchbot.json)
4Claude-UserAssistant, for a usercheck the address against the vendor’s IP list (https://claude.com/crawling/bots.json)
5Perplexity-UserAssistant, for a usercheck the address against the vendor’s IP list (https://www.perplexity.ai/perplexity-user.json)
6Amzn-SearchBotSearch crawlercheck the address against the vendor’s IP list (https://developer.amazon.com/amazonbot)
7ClaudeBotTraining crawlercheck the address against the vendor’s IP list (https://claude.com/crawling/bots.json)

Sources: Our agent list (Storewalk) and nginx: ngx_http_log_module

The AI agent traffic guide shows how to count them in Cloudflare and Google Analytics. A passive Storewalk scan does not measure traffic: its MEAS-001 check needs your own traffic logs, which a verified owner can upload in the dashboard. If agentic commerce itself is new to you, what is agentic commerce starts at the beginning, and we cover what stops an agent once it arrives separately.

Frequently asked questions

What does Adobe's 693.4% figure measure?

The rise in visits from generative AI tools to US retail sites over the 2025 holidays. It measures clicks, and Adobe says "the base of users remains modest".

How much do AI agents spend online?

No first-party figure we read measures it. Salesforce's $262 billion counts sales influenced by "AI and agents", recommendations included, and Morgan Stanley's $190 billion to $385 billion is a forecast for 2030.

Do shoppers from ChatGPT convert better?

Adobe says AI traffic converted 90% better than non-AI traffic in September 2026, and 9% worse in August 2025. That is across US retail sites in Adobe's data, not for any single store.

Are AI agents visiting websites more?

Cloudflare says "daily requests from AI agents on our network grew by more than 1,700%" over the year to September 2026. That counts requests on Cloudflare's network, not visits to any one site.

Where can I find reliable agentic commerce statistics?

On the publisher's own page, with its date and definition. If a statistic comes without both, treat it as a rumour with a percent sign. For your own store, your logs come first; after that, run our agent readiness scanner for what an agent meets when it arrives.

The sources below were last checked on the date shown.