Agentic commerce statistics
Sourced, dated figures on AI shopping agents and AI-referred traffic, plus our own measurements with their sample sizes. Each figure links to the organization that published it.
AI and shopping
Most of these figures measure the US market; read them as a leading indicator for other markets.
+138%US
Growth in AI-referred visits to US retail sites in May 2026 versus a year earlier.
Source: Adobe Analytics (Digital Commerce 360) ↗ ·
14xUS
AI-referred retail traffic since October 2024 (up 1,324%), across more than 1 trillion visits analyzed.
Source: Adobe Analytics (Digital Commerce 360) ↗ ·
+54%US
Higher conversion rate for AI-referred visits than for visits from other sources.
Source: Adobe Analytics (Digital Commerce 360) ↗ ·
51%US
Share of US consumers who used at least one AI tool to shop in the past month; the first time above 50%.
$3-5TGlobalForecast
McKinsey’s estimate of global agentic commerce by 2030 (up to $1 trillion in the US). A forecast, not a measurement.
Source: McKinsey (Digital Commerce 360) ↗ ·
Türkiye
Figures for the Turkish market. There is no official measurement of agent-driven sales in Türkiye yet.
94.5%Türkiye
Share of AI-driven website referrals in Türkiye that come from ChatGPT: the highest in the world (global average 80.9%). 39.7% of internet users aged 16 and over used ChatGPT in the past month.
Source: We Are Social and Meltwater, Digital 2026 (Turkish Minute) ↗ ·
₺4.57TTürkiye
Türkiye’s e-commerce volume in 2025; it grew 52% in nominal lira terms, with 634,000 businesses selling online. There is no official measurement of agent-driven sales in Türkiye yet.
Source: Turkish Ministry of Trade ↗ ·
The web and AI answers
How much traffic is non-human, and how stable AI recommendations are.
50%+Web
More than half of internet traffic now comes from non-human sources, for the first time.
Source: Cloudflare ↗ ·
<1%Web
Chance that two answers to the same recommendation question in ChatGPT or Google’s AI give the same list of brands; the same order comes back roughly once in a thousand. 600 volunteers, 12 prompts, 2,961 runs.
Source: SparkToro and Gumshoe ↗ · · 2,961 runs
Our own measurements
Measured by Specoria. We don’t name the stores we measured.
9 of 9Specoria
AI agents shopping in a real browser on five live stores: the two strongest models reached checkout on all 9 tasks that could be completed; two lighter models on 7 of 9. A small sample: read it as a field observation, not a success rate.
Source: Specoria field study · · 40 runs · 4 model families · 5 stores
836Specoria
Online stores measured for the Agent Readiness Index 2027 with the 12 checks of our free test’s first version: 365 Turkish and 471 English-language stores. The results are in the report (PDF); the sample is not random and its limits are set out there.
Source: Specoria, Agent Readiness Index 2027 · · 836 stores
Figures are published by the organizations named; only the last section is Specoria’s own data. Forecasts vary by organization.
Editorial policy
- We prefer the original publisher; where we cite a news report, the outlet is named in brackets.
- Every figure shows the date it was published and what it covers (US, Türkiye, global). Forecasts are labelled as forecasts.
- Our own figures carry the sample size and date, and say what the sample is not.
- We review the page every three months. A figure that is out of date or that we can no longer trace is removed, not kept.
- No figure on our site is made up or rounded up to sound bigger; an automated test in our code fails when a statistic on the site has no source.
What isn’t here yet
- Results of the Agent Readiness Index 2027 (average score, ready share, platform breakdowns): they are in the report PDF and will appear here once the figures are approved for the web.
- Sector figures for hotels and service businesses: we’ll publish them only once we have measured enough sites ourselves.
Run the free test on your store → Agent Readiness Index 2027 →