110+ x402 and Agentic Payments Statistics for 2026: Real Volume, Adoption and What AI Gets Wrong (Updated October)

Key takeaways

How much x402 activity is there?

Counts first. Every figure in this section is a transaction count or a cumulative dollar total, and most come either from Coinbase or from trackers reading the same contracts. They are accurate as counts. Read the next section before quoting any of them as demand.

Cumulative transaction counts

Coinbase reported 100 million x402 payments by the end of Q1 2026. This is Coinbase counting its own protocol, and the count includes test and self-payments. (Coinbase, May 2026)

Coinbase's developer docs say its facilitator alone has processed more than 100 million x402 transactions. The docs figure covers Coinbase's facilitator only, which is one reason it sits below all-facilitator trackers. (Coinbase docs, 2026)

Coinbase said x402 had handled 160 million+ agentic payments in a year. An executive post rather than a filing, so treat it as the top of the vendor range. (Coinbase on X, Jun 2026)

At Agentic.Market's April launch, Coinbase cited 165 million x402 transactions, $50 million in volume and 69,000 agents, an average of $0.31 per transaction. Divide one by the other and the average payment is about 31 cents, before any filtering. Another report of the same launch gave 480,000 agents, so the agent count is the least reliable of the three. (Coinbase via Yahoo Finance / Dealroom, Apr 2026)

By mid-February 2026, x402 had passed 120 million transactions and $41 million in cumulative payments across chains, with Solana at 45 million and $16.4 million. Cumulative totals only ever rise, so they can look healthy while weekly activity falls. Solana's weekly count dropped by more than 90% over the same weeks. (SolanaFloor via Phemex, Feb 2026)

Chainalysis confirmed cumulative agentic transactions on Base passed 100 million within roughly three quarters of near-zero activity. This is the independent check on Coinbase's 100 million. It confirms the count. It says nothing about whether those payments bought anything. (Chainalysis, Jun 2026)

An academic count found 136.7 million x402 settlements worth $44.1 million gross on Base, and 49.5 million worth $9.0 million on Solana, over 280 days. The gross totals look close to Coinbase's. The same paper then shows how little of that value traces to a genuine service. (Ling et al., arXiv, Jul 2026)

A separate measurement counted about 119 million x402 transactions (91.5M Base, 28.2M Solana) from October to December 2025. A second academic team, on a different window, landed in the same range. Counting methods agree; the argument is about what the counts mean. (Wang et al., arXiv, Jul 2026)

Base (9.7 million) and Polygon (8.1 million) carried almost all of 18.3 million x402-style payments in August 2026. Polygon runs almost level with Base here, which will surprise anyone who has only read Coinbase's '90% on Base' line. Different month, different counting scope. (Bitquery, Sep 2026)

What the payments look like

95% of x402 value now comes from payments of $1 or more, up from 49% in early 2025; payments of 10 cents to $1 fell from 46% to 4%. Value has shifted to larger payments, which cuts against the sub-cent micropayment story most explainers still tell. This is share of value, so tiny payments can still dominate the count. (Chainalysis, Jun 2026)

Conversion from x402 testers to paying users improved fourfold over six months. No base rate was published, so you can't turn this into a number of paying users. (Chainalysis, Jun 2026)

x402 wallets average 197 days old against 423 days for typical Base wallets, and hold 26 tokens against 4. Young, busy wallets holding lots of tokens fit bots and test rigs as well as they fit real agents. (Chainalysis, Jun 2026)

Of 25,163 x402 resources listed in Coinbase's Bazaar catalogue, only 249 payee addresses ever earned $10 or more. Marketplace size is a poor proxy for a working economy: 25,163 listings, 249 payees who ever earned ten dollars. (Ling et al., arXiv, Jul 2026)

One Bazaar-listed x402 research API had received 8 paid calls in its first two weeks on mainnet, all from auditors testing the endpoint and none from an agent doing real work. A merchant's view of the same gap the on-chain audits show: the payment rail works, the agent demand hasn't arrived. (Elements Live, 15 Oct 2026)

"We've had an x402 research API live on mainnet since 1 October, listed in the Bazaar. So far every paid call has come from auditors testing the endpoint, not from an agent doing real work."
Marko Milojevic, founder of Elements Live (elements-live.com), as of 15 October 2026

Only 52% of 910 x402 hosts tested returned a working payment challenge. Nearly half the advertised paid endpoints didn't return a usable payment request when tested. 'x402-enabled' on a listing doesn't mean you can pay it. (Ling et al., arXiv, Jul 2026)

x402 activity is extremely concentrated: the payer Gini coefficient on Base was 0.987 and the largest single recipient took 23% of settlements. A Gini of 0.987 means a handful of payers do almost all the paying. If one operator stops, the headline number moves with it. (Ling et al., arXiv, Jul 2026)

Why the x402 numbers don't agree

In one week in October 2026 I found x402 described as a $24 million-a-month market, a $41.9 million cumulative one, and a market where AI agents moved about $57,000 in a month. Each figure was published in good faith. They count different things, over different windows, through different facilitators, and only some remove wash trading.

The table below lines up the figures you are most likely to see quoted. Bookmark it; the link goes straight here.

Wash trading and operator-internal flows

Bloomberg's widely quoted $24 million monthly x402 volume (Feb 5 - Mar 7, 2026) drops to about $3 million on Allium's on-chain count and $1.6 million after removing wash trades. Same 30 days, three numbers. The $24 million version is still the one most articles, and most AI answers, repeat. (a16z crypto, Mar 2026)

Wash-trade filtering cut one month of x402 volume by about 93%, a roughly fifteenfold gap from the headline figure. A fifteenfold gap between headline and adjusted volume is the single most useful fact on this page for anyone writing about x402. (a16z crypto / ForkLog, Mar 2026)

85% of x402 settlements on Base were operator-internal: 21.2% provably fictitious and 63.8% moving within linked wallet clusters. The authors' robustness range runs from 72.4% to 85%, so even the cautious end puts most activity inside wallets the operators control. (Ling et al., arXiv, Jul 2026)

Provably fictitious settlements were only 21% of x402 transactions but 54% of the dollar value. The fake payments were bigger than the real ones, which is why dollar volume is inflated more than transaction counts. (Ling et al., arXiv, Jul 2026)

Only $187,861 of $44.1 million in x402 volume could be traced to a nameable paid service; at most $20.3 million was not provably manufactured. $187,861 is a floor, because some genuine payments can't be matched to a named service, and $20.3 million is the ceiling. My read of the paper puts the truth nearer the floor. (Ling et al., arXiv, Jul 2026)

The authors call settlement count a 'Goodhart metric' that measures the incentive to transact rather than real adoption. The plainest explanation of why counts rose while value didn't: projects were rewarded for transacting, so they transacted. (Ling et al., arXiv, Jul 2026)

Facilitators paid an estimated $356,000 in gas to process x402 settlements, more than the provable genuine payments. If the estimate holds, facilitators spent more on gas than the provably genuine payments were worth. A free tier is a subsidy, and subsidies attract volume. (Ling et al., arXiv, Jul 2026)

Artemis classified an average 63.94% of x402 transactions as 'gamed' between 9 November 2025 and 7 February 2026, peaking at 90.21% on one day. Artemis doesn't publish its heuristics, so treat 63.94% as one firm's classification. It sits in the same band as the academic estimates. (Four Pillars, Feb 2026)

99.99% of x402 transactions on Polygon in late January 2026 sent 0.01 USDC to one address, and 30 new wallets made 91.8% of 2.1M transactions. One cent to one address, over and over, from 30 new wallets. This is what pushed Polygon's share up in January, and it stopped on 1 February. (Four Pillars, Feb 2026)

A single meme coin's pay-to-mint mechanic drove a sharp, short-lived spike in x402 transaction counts. Pay-to-mint traffic counts as an x402 payment, so a meme coin launch can move the totals for weeks. (Chainalysis, Jun 2026)

Adjusted totals and the lowest credible figures

Visa and Artemis put cumulative x402 activity at about $15.0 million and 109.6 million transactions after removing identified wash and test activity. Visa funded this work and runs competing products, but its filtered total lines up with the independent papers better than any vendor figure does. (Visa and Artemis, Jul 2026)

TRM Labs counted about $52.7 million settled through known x402 facilitators across 198.9 million transactions since May 2025, on Base, Solana and Polygon; 99.6% of the value was USDC. TRM's gross total sits between Visa's adjusted $15.0 million and Coinbase's $53 million claim. USDC carries almost all of it. (TRM Labs, Sep 2026)

After removing self-payments, bulk flows and sellers with few buyers, TRM kept $25.62 million of x402 commerce and judged only 0.6% to 7.5% of it agentic, a run-rate of roughly $5,000 to $11,000 a month. The lowest credible measure of real agent spending yet published. TRM's test looks for varied payments to several sellers, so it may miss agents that only ever buy one thing, and TRM says it can't fully separate agents from scripts. (TRM Labs, Sep 2026)

About 4,000 wallets drive roughly 90% of adjusted x402 spending. If 4,000 wallets account for nine-tenths of adjusted spend, a few large users can swing monthly totals on their own. (Visa and Artemis via The Defiant, Jul 2026)

One on-chain count put x402-style payments at $2.59 billion in August 2026, but about 88% of that came from a single bridging contract on Arbitrum. Strip out that one contract and roughly $317 million remains, still far above every other estimate, so 'x402-style' in this dataset catches much more than agent payments. (Bitquery, Sep 2026)

Bitquery estimated agent payments on Base and Polygon at only about $57,000 in August 2026. The agent classification is Bitquery's own inference. It is the lowest credible monthly figure we found, and the one Google's AI Overview quoted. (Bitquery, Sep 2026)

Solana's weekly x402 transactions fell more than 90%, from about 6.8 million in December 2025 to under 510,000 in the week to 9 February 2026. Cumulative totals hid this fall completely. Weekly data showed it straight away. (SolanaFloor via Phemex, Feb 2026)

Allium's x402 daily metrics table counts transactions, volume, unique senders and receivers by facilitator, but documents no wash-trade filter. Fine for trends. With no documented wash filter, its volume column should be read as gross. (Allium, 2026)

Trackers put cumulative x402 transactions anywhere from 120 million to 187 million depending on which one you check. A spread of 67 million transactions between trackers, and no agreed reference. Name the tracker and the window every time you quote a count. (Crypto Briefing, Sep 2026)

What AI assistants say about x402

We asked ChatGPT, Perplexity and Google's AI Overviews 14 questions that a developer, founder, finance lead, journalist or sceptic would actually type. Every run was made from a UK location on 13 and 14 October 2026: ChatGPT signed in using Temporary Chat, Perplexity signed in with its incognito setting, and Google AI Overviews signed out in a private window. Across the 42 answers we logged each company, protocol, link and figure: 465 records in all. The full workbook is available to download at the end of this page.

Same assistants, same day, two different x402s

Running Q6 and Q7 on the same day was the moment the study paid for itself. Ask whether x402 is ready and you get vendor counters. Ask how much it is really used and the same assistants turn sceptical and quote adjusted data.

Our data Asked whether x402 is production-ready, all three AI platforms said yes and none mentioned wash trading; two quoted cumulative counts of 100M+ and 191M transactions. 'Is it ready?' questions pull vendor docs and cumulative counters, the least filtered numbers available. (cryptocontent.dev study, Oct 2026)

Our data Asked how much real usage x402 has, all three AI platforms put organic payment value at around $1M a month or less (ChatGPT: under $1M; Perplexity: $0.3M-1.3M; Google: ~$57K of agent volume in August 2026). Ask about real usage and you get adjusted figures; ask about readiness and you don't. Both answers came from the same assistants on the same day. (cryptocontent.dev study, Oct 2026)

Our data Of 14 x402 volume figures quoted across 42 AI answers, 11 were adjusted figures and 3 were unadjusted gross totals. Most figures AI quotes are now adjusted ones. The unadjusted ones turn up where they do the most damage, in answers about whether to build. (cryptocontent.dev study, Oct 2026)

Our data a16z crypto's adjusted figure ($1.6M a month against a $24M headline) appeared in 1 of 42 AI answers. The best-known correction in crypto media barely reaches AI answers. Perplexity surfaced it only when asked which data sources to trust. (cryptocontent.dev study, Oct 2026)

Who gets named, and who gets the link

Our data x402 was named in 30 of 42 AI answers (71%) to buyer-style questions about agentic payments. x402 is the default answer to developer-framed questions. It was missing from most answers about Shopify stores, finance controls and UK providers. (cryptocontent.dev study, Oct 2026)

Our data Coinbase's own website was cited in 5 of 42 AI answers (12%), even though Coinbase created x402; Stripe was named in 26 of 42. AI now credits the protocol to the x402 Foundation and quotes Finextra, privacy.com or Instagram posts when describing Coinbase's products. (cryptocontent.dev study, Oct 2026)

Our data ChatGPT linked to the named company's own site 65 times and used unnamed links 9 times; Google AI Overviews did the reverse, with 33 own-site links and 55 unnamed links to blogs, LinkedIn, Instagram and YouTube. If you want Google's AI Overview to describe your product accurately, the page it quotes is more likely to be a blog post or a LinkedIn update than your docs. (cryptocontent.dev study, Oct 2026)

Our data Alchemy's own comparison page was the main source for the 'which wallet lets an AI agent hold funds' question on 2 of 3 AI platforms, and Alchemy appeared in both answers. Vendor comparison pages set the shortlist. Crossmint's own comparison post did the same job in Google's answer, and a commerce platform's stats guide drove two of the 'is it hype?' answers. (cryptocontent.dev study, Oct 2026)

Where the answers split or went wrong

Our data The same protocol-choice question got three different first recommendations: x402 on ChatGPT, ACP on Perplexity and AP2 on Google AI Overviews. For a product manager choosing a protocol, the assistant they happen to use decides the recommendation. (cryptocontent.dev study, Oct 2026)

Our data The three AI platforms agreed on the top recommendation for 2 of 14 questions (x402 with USDC on Base; Cloudflare and TollBit for publishers). Agreement came only on narrow questions with an obvious incumbent. Broad strategy questions split three ways. (cryptocontent.dev study, Oct 2026)

Our data 2 of 3 AI platforms implied UK businesses can use Stripe to accept stablecoin payments; Stripe's stablecoin terms (updated 16 April 2026) cover the US, EEA, Switzerland, Mexico and Hong Kong. We checked Stripe's published terms after both answers told a UK company it could sign up. ChatGPT, citing Stripe's support pages, had it right. (cryptocontent.dev study, Oct 2026)

Our data None of ChatGPT, Perplexity or Google AI Overviews mentioned x402 when asked 'Is agentic commerce actually happening, or is it mostly hype? Give me the real numbers'. AI treats 'agentic commerce' as a retail shopping topic. Anyone sceptical about x402 has to ask about x402 by name to see its numbers. (cryptocontent.dev study, Oct 2026)

Which chains carry x402, and what it costs

Base carries most x402 value; Solana carries a large share of the count; Polygon and the XRP Ledger have both had bursts. Chain shares in press reports swing wildly because they mix facilitators, windows and count with value.

Chain and facilitator share

Coinbase's facilitator handled 63.5% of x402 settlements on Base (86.9 million, $27.9 million). Coinbase's own facilitator is the biggest on Base but handles under two-thirds, so Coinbase-only figures miss a third of the chain. (Ling et al., arXiv, Jul 2026)

Coinbase says more than 90% of x402 volume runs on Base. This is Coinbase's facilitator view. All-facilitator trackers that include Solana report much lower Base shares. (Coinbase, May 2026)

Over 93% of x402 sellers rely on a single facilitator. Single-facilitator dependence matters more once you read the security results in the risks section: every major facilitator tested had flaws. (Wang et al., arXiv, Jul 2026)

The XRP Ledger processed more than 8 million x402 payments in 30 days, but its all-time settled value was 11,318 RLUSD and 6,928.5 XRP. The same shape as Base and Polygon, only sharper: millions of payments, settled value in the low thousands of dollars. (XRPL AI Hub via KuCoin, Oct 2026)

Costs and service levels

Coinbase's x402 facilitator gives the first 1,000 settlements each month free, then charges $0.001 per settlement, with no chargeback fees. On a one-cent payment, a $0.001 fee is 10% of the sale. Per-request pricing below a cent rarely works without batching or prepaid credit. (Coinbase docs, 2026)

Coinbase targets p95 settlement of 900ms on Base and 1,500ms on Solana for x402, with 99.9% availability. These are published targets, not measured averages. (Coinbase docs, 2026)

Coinbase's facilitator supports x402 on Base, Polygon, Arbitrum, World and Solana mainnets. Five mainnets on one facilitator. The XRP Ledger and others run through separate facilitators with their own counts. (Coinbase docs, 2026)

Who's adopting x402 and agent payments

Sign-ups are easy to count. Payments are harder. Most adoption figures in this space measure partners, registrations or capability, and very few measure money moving.

Networks, platforms and members

Visa says more than 100 partners are working with Visa Intelligent Commerce, 30+ building in its sandbox and 20+ agent builders integrating directly. Partner counts measure who has signed up to build, not who is taking payments. (Visa, Dec 2025)

Visa reported only 'hundreds' of completed agent-initiated transactions by December 2025. Hundreds of agent card purchases against x402's hundred million settlements. They measure different things: Visa's are card purchases by agents, x402's are mostly machine-to-machine settlements. (Visa, Dec 2025)

Ripple joined the x402 Foundation as a premier member alongside Visa and Mastercard in July 2026. Ripple's membership brought the XRP Ledger and its RLUSD stablecoin into the x402 group. (247 Wall St, Jul 2026)

Almost none of 20 large B2B payment and procurement tools let an AI agent move money; Ramp, Coupa, Melio and BILL lead. B2B is where most agent spending is forecast to happen, and the tools finance teams use mostly can't let an agent pay yet. (Flagship Advisory Partners, Aug 2026)

Alipay's AI Pay passed 120 million transactions in a single week in February 2026, through agents in apps, smart glasses and the Qwen chatbot. One week of Alipay AI Pay is close to x402's entire cumulative count, on ordinary fiat rails. Ant Group doesn't define what makes a payment 'AI', so use it as scale context only. (Alipay Ant Group, Feb 2026)

Publishers and HTTP 402

Cloudflare says its customers already send more than one billion HTTP 402 'payment required' responses on an average day. 402 is the status code x402 is built on. Most of these are refusals to crawlers rather than payment requests that got paid, but the plumbing for pay-per-request already runs at that scale. (Cloudflare, Aug 2025)

OpenAI's crawlers made up to 3,700 page requests for every visitor they referred back at the March 2025 peak; Perplexity settled below 200:1. Crawl-to-referral ratios this lopsided are why publishers want crawlers to pay, and why Cloudflare built pay per crawl on HTTP 402. (Cloudflare Radar via Leadgen Economy, May 2026)

Agent identity and Know-Your-Agent

Before an agent can pay, someone has to know what it is and what it's allowed to spend. Two separate answers are being built. Crypto has on-chain registries such as ERC-8004. Card networks and wallet providers are building Know-Your-Agent (KYA) checks, where the payment provider verifies the agent and its permissions before money moves. So far the two stacks don't reference each other.

186,092 AI agents were registered under the ERC-8004 identity standard across 26 chains by May 2026, four months after it went live. Registrations, not active agents. A wallet can register an identity in one transaction and never use it again. (8004scan via Hashlock, May 2026)

BNB Smart Chain hosted 79,260 ERC-8004 agents and Base 28,430, ahead of Ethereum's 15,401. BNB Chain leads registrations even though Base carries most x402 payments. Identity and payments are separate layers with separate leaders. (8004scan via Hashlock, May 2026)

Ant International, Mastercard and Visa began work on a shared Know-Your-Agent interoperability framework in September 2026, with each network keeping its own verification and decisioning. Three competing agent-identity systems are agreeing common principles rather than one standard. None of the three is built on ERC-8004. (Ant International, Sep 2026)

Eleven firms contributed to the Monetary Authority of Singapore's SAFR (Safeguards for Agentic Finance at Runtime) framework, among them Ant International, Circle, Google, Mastercard, Visa, DBS, HSBC and J.P. Morgan. SAFR reviews and logs each action an agent proposes before it runs. Circle issues USDC, x402's main settlement asset, so the stablecoin side helped write it too. Version 1.1 was published on 9 October 2026. (MAS, Oct 2026)

The agentic payments protocols

Seven protocol families compete for the same job, and most AI answers we logged recommended combining two or three of them. The table gives the short version. The last column lists the numbers that would show whether each one has real use; almost none of them have been published. The figures below show how little volume any protocol outside x402 has reported. Ant International's AMP is the newest entry and the only one built on mobile wallets.

Launches and membership

The x402 Foundation launched under the Linux Foundation with 40 member organisations, including Visa, Mastercard, American Express, Stripe, Google, AWS, Shopify and Coinbase. The foundation now owns the specification, which is why AI answers increasingly credit it rather than Coinbase. (Linux Foundation, Jul 2026)

17 premier members back the x402 Foundation, among them both major card networks and the Solana, Stellar and Monad foundations. Card networks backing a stablecoin protocol is the clearest sign the two kinds of rail are expected to sit side by side. (Linux Foundation, Jul 2026)

Google launched the Agent Payments Protocol (AP2) on 16 September 2025 with more than 60 partners, including Mastercard, American Express, PayPal, Coinbase and Revolut. AP2 is an authorisation layer that can sit over cards or stablecoins. Google has discussed adding x402 for the stablecoin leg. (MediaPost, Sep 2025)

Google's Universal Commerce Protocol was co-developed with Shopify, Etsy, Wayfair, Target and Walmart and endorsed by 20+ global partners. UCP is retail-led, with merchants rather than payment firms as co-developers. (InfoQ, Jan 2026)

OpenAI and Stripe released the Agentic Commerce Protocol with ChatGPT Instant Checkout, starting with Etsy and then Shopify merchants. ACP runs on card rails through Stripe's Shared Payment Tokens, the opposite design choice to x402. (Shopifreaks, Oct 2025)

Ant International open-sourced the Agentic Mobile Protocol (AMP) in April 2026 and says its agent-to-agent settlement can handle payments as small as $0.000001. A millionth of a dollar, settled on fiat wallet rails. That is the pay-per-request job x402 is known for, from a payments group with no stablecoin in the stack. Ant hasn't published any AMP volume, so read it as a design claim. (Ant International, Apr 2026)

AMP's first rollout phase covers 10 Alipay+ wallets serving 1.5 billion user accounts, plus seven acquirers: Adyen, Allinpay, Checkout.com, Fiserv, Global Payments, Nuvei and Worldline. On paper that reach beats every protocol in the table except the card networks. Like Visa's partner count, it measures who can pay through AMP, not who does. (Ant International, Sep 2026)

Ant says AMP halves the number of steps needed to link a payment agent to a wallet, compared with card-binding. Ant's own measure, with no baseline published. (Ant International, Apr 2026)

Volume and results

Stripe and Tempo's Machine Payments Protocol settled about $25,000 across roughly 115,000 transactions in its first few weeks. Small, but published with a dollar figure next to the count, which is more than most protocols offer. (Visa and Artemis, Jul 2026)

Stripe announced 288 products and features at Sessions 2026, including the Machine Payments Protocol and agent wallets, without publishing any agentic volume figures. 288 launches and no agentic volume figure. When a payments company this size publishes no number, assume the number is small. (Stripe, Apr 2026)

Walmart put about 200,000 products into ChatGPT Instant Checkout; in-chat purchases converted at one-third the rate of shoppers who clicked through to Walmart's site. The clearest field result we found for agent checkout: shoppers wanted to see the product page before paying. (Search Engine Land, Mar 2026)

OpenAI began phasing out ChatGPT Instant Checkout in March 2026, moving purchases to merchant sites and an apps format, about six months after launching it with the Agentic Commerce Protocol. Six months from launch to retreat. OpenAI kept the protocol and sent the purchase back to merchant sites. (ForkLog, Mar 2026)

Stablecoins as the payment rail

x402 settles almost entirely in USDC, so the stablecoin market sets its ceiling. The same gap between raw and real volume shows up here, at a much larger scale.

Supply

Total stablecoin supply hit a record $315 billion in Q1 2026, with USDC at about $78 billion. Supply moves every week, so date any figure you quote. (Cryptonews via Yahoo Finance, Apr 2026)

Coinbase held a record $17.8 billion of USDC in its products in Q4 2025, out of an average $76.2 billion USDC market cap. Coinbase earns income on USDC held in its products, which gives it a direct interest in USDC payment volume. (Coinbase, Feb 2026)

Coinbase held an average $19 billion of USDC in its products in Q1 2026, more than 25% of all USDC in circulation. Taken from a reported earnings deck and not yet checked against the filing. (Coinbase, May 2026)

Raw volume vs real payments

Raw on-chain stablecoin volume runs up to $35 trillion a year, but McKinsey and Artemis estimate only about $390 billion of it is genuine payments. About 1% of raw stablecoin volume is real payments. x402 has the same problem in miniature. (McKinsey and Artemis, Feb 2026)

Genuine stablecoin payments are about 0.02% of global payments volume. Stablecoin payments are a rounding error in global payments, and agent payments are a slice of that slice. (McKinsey and Artemis, Feb 2026)

B2B makes up about 60% of real stablecoin payments, roughly $226 billion a year, up 733% year on year. Real stablecoin growth is in B2B, which is also where the largest agent-spending forecasts point. (McKinsey and Artemis, Feb 2026)

Stablecoin-linked card spending reached $4.5 billion in 2025, up 673% on 2024. Cards funded by stablecoins grew fast, so many 'stablecoin payments' still end on card rails. (McKinsey and Artemis, Feb 2026)

Asia originates about 60% of real stablecoin payments (~$245 billion), against $95 billion from North America and $50 billion from Europe. Europe trails on real stablecoin payments. UK and EU businesses reading US-centric x402 coverage should expect fewer local providers. (McKinsey and Artemis, Feb 2026)

Stripe's stablecoin rollout

Stripe added stablecoin payment acceptance in 32 more markets and stablecoin-backed cards in 30 countries at Sessions 2026. Acceptance is market by market. The UK is not in Stripe's published stablecoin terms, which cover the US, EEA, Switzerland, Mexico and Hong Kong (checked 14 October 2026). (Stripe, Apr 2026)

Stripe's global payouts in stablecoins were previewed for 160 countries. Payouts reach far more countries than acceptance does. (Stripe, Apr 2026)

Forecasts and demand, and how far apart they are

Here's the uncomfortable bit. The 2030 forecasts run to trillions. Today's measured demand is a single-digit share of shoppers willing to let an agent pay on its own. Both can be honest numbers. They aren't measuring the same thing, and none of the forecasts below is built on x402's adjusted volume.

What consumers say

Only 7% of US and UK adults would let an AI assistant buy for them autonomously, even under preset conditions; 53% are uncomfortable with it buying on their behalf. Seven percent. Every 2030 forecast in the table assumes that number grows several times over. (ACI Worldwide / YouGov, Sep 2026)

20% of consumers accept AI shopping recommendations but not autonomous purchases, and 14% want to approve every purchase manually. Most people want an agent that recommends and then asks. That suits authorisation layers like AP2 better than fully autonomous payment. (ACI Worldwide / YouGov, Sep 2026)

Consumers in a seven-market survey expect AI agents to make 15% of their purchases within five years, up from 9% a year earlier. An expectation, not behaviour, and well above the 7% who would let an agent buy on its own today. The survey covered 8,027 consumers in the US, UK, France, Brazil, China, Singapore and Australia; the release doesn't say whether 15% is a seven-market figure or a US one. (Global Payments, Sep 2026)

81.3% of travellers in a nine-market survey already use AI tools to explore destinations, but only 26% are interested in letting AI move closer to payments and decisions. Interest falls by more than two-thirds between finding and paying. That gap is what Know-Your-Agent checks and wallet-side approval are meant to close. (Alipay+ and S&P Global, Sep 2026)

55% of US and UK shoppers aren't comfortable with AI agents making purchases for them, down from about 7 in 10 who were at least somewhat comfortable in Q4 2025. In this tracker comfort is falling, though the sample size isn't published. (Riskified, Apr 2026)

19% of US and UK consumers already use AI agents to deal with brands, and Braze expects that to reach 46% by the end of 2026. The 46% is Braze's projection. The measured figure is 19%. (Braze, Feb 2026)

27% of consumers refuse to share any data with AI agents, even for a better experience. An agent that can't see your history can't shop well for you, so data refusal caps how useful agents get. (Braze, Feb 2026)

47% of shoppers across 12 markets use AI tools for at least one shopping task. Using AI for one shopping task is a low bar: comparing two products counts. (Visa / Morning Consult, Dec 2025)

Nearly 23% of Americans made at least one AI-assisted purchase in the past month. 'AI-assisted' covers research and recommendations, so this doesn't measure agents paying. (Morgan Stanley AlphaWise, Dec 2025)

44% of people who have tried AI-powered search now call it their preferred way to search, against 31% for traditional search. Search preference, not purchasing. It does show where buying journeys now start. (McKinsey, Oct 2025)

What merchants and networks say

Nine in ten merchants say they are actively preparing for agentic commerce, and 42% are already testing it. Preparation is cheap. Payment volume is the number to watch. (Checkout.com, 2026)

Checkout.com data shows about 3% of transactions currently involve AI agents, while 89% of merchants are preparing. The source doesn't say which transactions the 3% covers. Google's AI Overview restated it as transactions 'fully executed by an AI agent', which the source doesn't claim. (Checkout.com via Forkast, Jul 2026)

AI-driven traffic to US and EU retail sites grew 138% year on year in May 2026, and AI-referred retail visitors produced 53% more revenue per visit. AI-referred shoppers spend more per visit, and they still check out on the merchant's own site. (Flagship Advisory Partners, Aug 2026)

Mastercard projects that about one in ten consumers will routinely use AI agents to make purchases by 2030. One in ten by 2030, from a card network building the product. Read it as a target. (Mastercard, Sep 2026)

Juniper Research expects $8 billion of agentic commerce transaction value in 2026, rising to $1.5 trillion of spending by 2030. Getting from $8 billion to $1.5 trillion in four years needs growth of roughly 270% a year, and the two figures may not measure the same thing. (Juniper Research via TechNode, Apr 2026)

Rules, security and fraud risk

Two kinds of risk matter to anyone accepting agent payments: the rules on the stablecoin itself, and the security of the software that verifies and settles each payment. The second gets far less coverage.

Facilitator security

All 15 major x402 facilitators violated at least one of eight security rules, with 31 previously unknown exploitable attacks found. Every facilitator tested broke at least one rule. If you accept x402, your facilitator's security is part of your payment risk. (Wang et al., arXiv, Jul 2026)

Attacks included 'free shopping' (10 high-risk cases), asset theft at a top-volume facilitator and gas abuse. 'Free shopping' means a buyer received the paid resource without a valid payment ever settling. (Wang et al., arXiv, Jul 2026)

Coinbase, PayAI and Mogami confirmed six vulnerabilities after disclosure; one facilitator had no security reporting channel. Quick fixes from the largest facilitators are the good news here. (Wang et al., arXiv, Jul 2026)

Fraud and liability

53.9% of shoppers think AI agents could increase online fraud risk and 50.8% say AI platforms should be liable for unauthorised purchases. x402 payments are final, with no chargeback, so on stablecoin rails the merchant and payer carry this risk themselves. (Riskified, Apr 2026)

73.9% of shoppers expect strong safeguards such as biometrics or one-time passwords before an agent can buy. Card networks are building agent authentication for exactly this reason. x402 leaves authentication to the wallet. (Riskified, Apr 2026)

Half of Americans (50%) are concerned about the security of their payment details when AI agents shop, and a third of consumers want to approve every transaction before it is paid. A third is more than double ACI's 14%; the two surveys used different wording and markets. Either way, a large group wants the final say kept with the wallet, which is how AMP and AP2 are designed. (Global Payments, Sep 2026)

US and UK rules

The US GENIUS Act became law on 18 July 2025, requiring at least $1 of permitted reserves for every $1 of payment stablecoins issued. Full reserve backing under GENIUS is the basis for USDC's role as x402's default asset. (Congressional Research Service, Aug 2026)

Stablecoin issuers above $10 billion outstanding must be federally supervised; those above $50 billion must publish audited annual financial statements. Almost all x402 value settles in one stablecoin issued well above both thresholds. (Congressional Research Service, Aug 2026)

US regulators missed the GENIUS Act's one-year deadline of 18 July 2026 for final stablecoin rules; the Act takes effect by 18 January 2027 at the latest. Missing the deadline delays the detail, not the law. The latest effective date is fixed. (Crowdfund Insider, Jul 2026)

The FCA published final rules for the UK cryptoasset regime on 30 June 2026; the regime starts on 25 October 2027. The new UK regime is a year away. Until then, UK merchants should check the FCA's current guidance on accepting cryptoassets before signing up to a provider. (AIMA, Jun 2026)

UK firms can apply for FCA crypto authorisation between 30 September 2026 and 28 February 2027. The application window closes eight months before the regime opens. (AIMA, Jun 2026)

What these statistics mean in practice

If you're quoting x402 volume

Use an adjusted figure and give its window: Visa and Artemis's $15.0 million to April 2026, or a16z's $1.6 million for February to March 2026. If the question is how much AI agents themselves spend, TRM Labs' $5,000 to $11,000 a month is the figure to quote. Never present a cumulative transaction count as adoption. If you need one line, 'over 100 million transactions, but genuine payment value in the low millions of dollars' is accurate and defensible.

If you're building on x402

• Price above a cent or batch payments; a $0.001 settlement fee eats 10% of a one-cent sale.
• Ask your facilitator which of the published vulnerabilities it has fixed, and whether it has a security contact.
• Test your own endpoint's payment challenge. Nearly half of listed hosts failed that check.
• Plan for no chargebacks: refunds are your policy, not the network's.

If you sell into this market and want AI to recommend you

The vendors named most often in our 42 answers weren't the biggest. They published comparison pages and statistics that AI could quote: Alchemy, Crossmint and Nevermined all showed up this way. Coinbase, which built the protocol, was cited from its own site in 5 answers. Dated, sourced numbers on your own site get you into answers; product pages alone mostly don't. Our 2026 AI Citation Visibility Study found the same pattern across 50 crypto protocols.

If you're a UK business

Two of three AI assistants told our UK test business it could sign up for Stripe's stablecoin acceptance. Stripe's terms say otherwise. Check a provider's published terms for your country before building, and watch the FCA regime that starts on 25 October 2027.

What nobody has measured yet

Builders working on agent payments keep asking the same ten questions. We checked each one against the sources on this page. One has a solid published answer, four have partial answers, and five have no public data at all. The status reflects what we could find as of 19 October 2026.

Frequently asked questions

Methodology

We checked 142 candidate statistics between 10 and 14 October 2026 and kept 114. The other 28 were rejected. The main reasons: no traceable origin (figures that appear only in a headline, an X post or a LinkedIn update), inconsistency with every on-chain count (a $600 million annualised x402 figure, for example), superseded editions, and a few cut to stop one publisher dominating the page. One first-hand merchant figure was added on 16 October 2026.

On 18 October 2026 we checked 11 more candidates on Ant International's Agentic Mobile Protocol, Know-Your-Agent identity and consumer trust, and added 8. The same day we added two figures from TRM Labs. We had rejected TRM's x402 total on 14 October because we couldn't find TRM's own publication; we have since read it on TRM's blog. Each was read at Ant International's, MAS's or Global Payments' own release or at a named repost of it. Where a secondary write-up disagreed with the release, we used the release: AMP's first phase has seven external acquirers, not eight, and Global Payments' 50% security figure covers Americans only.

On 19 October 2026 we added the Open questions section and a "What to measure" column to the protocols table. The section reuses figures already on this page plus one HUMAN Security blocking figure, which isn't counted in the 125 statistics. An unconfirmed BrightEdge estimate of merchants allowing agent checkout, reported by the Wall Street Journal, was left out until it can be read at source. Later the same day, Benji Fisher of UCP Checker confirmed its October census figures for this page; they appear in questions 3 and 6 and aren't counted in the 125 statistics.

Every kept statistic was read at its source or at the nearest named report of it. Each carries a confidence tier in our ledger:

• Strong (77): primary source read, with figure, sample and date confirmed.
• Moderate (41): a named secondary report of a primary source, or a vendor's own figure.
• Weak (7): source unnamed or in conflict with a stronger figure. Shown only in the reconciliation table and the chain-share note, and never in key takeaways.

Our own data (12 statistics). The x402 AI Answer Study ran 14 buyer-style questions on ChatGPT, Perplexity and Google AI Overviews on 13 and 14 October 2026, from the UK. One person ran and logged every answer. We recorded 465 records: each named company or protocol, its link status (own site, third-party link, named without a link, or linked without being named) and every figure quoted, classed as gross volume, adjusted volume, transaction count, forecast or other. Answers vary by account, location and day, and several answers showed signs of personalisation (GBP prices, UK spelling, references to a marketing use case), so treat the counts as a dated snapshot.

Most people who ask an AI assistant a question are signed in, so signed-in answers are closer to what a real buyer sees. Signing out and using private browsers is mostly something SEO and GEO researchers do. The account behind these runs belongs to a UK content and research business, which explains the marketing references; a developer or trader asking the same questions may get different detail.

All charts are drawn from the numbers in our ledger; none are copied from the original publishers. Figures from different studies are never plotted on the same axis unless they share a definition.

Next refresh: 14 January 2027. x402 volume figures and stablecoin supply date fastest; the Visa and Artemis report, Chainalysis's x402 analysis and the FCA timetable are the sources most likely to publish updates before then.

Sources

Original data: cryptocontent.dev x402 AI Answer Study, October 2026 (this page). Third-party sources, in order of first citation, with sources added on 18 and 19 October 2026 listed last:

1. Marko Milojevic, Elements Live, merchant data supplied to the author via LinkedIn (15 Oct 2026)
2. a16z crypto, The honest number behind AI agent payments (Mar 2026)
3. Ling et al., arXiv, How Agentic Is Agentic Commerce? A Population-Scale Measurement of x402 Adoption and Authenticity (Jul 2026)
4. Visa and Artemis, Agentic Payments from the Ground Up (Jul 2026)
5. Search Engine Land, Walmart: ChatGPT checkout converted 3x worse (Mar 2026)
6. Coinbase, Q1 2026 earnings disclosure (via The Defiant / KuCoin) (May 2026)
7. Coinbase docs, x402 FAQ (2026)
8. Coinbase via Yahoo Finance / Dealroom, Coinbase launches AI app store (Apr 2026)
9. SolanaFloor via Phemex, Solana's x402 market share drops (Feb 2026)
10. Chainalysis, Agentic payments on Base (via Crowdfund Insider) (Jun 2026)
11. Wang et al., arXiv, When HTTP 402 Meets the Blockchain: Risks on Emerging x402 Payments (Jul 2026)
12. Bitquery, x402 Protocol: $2.6B a Month Across 5 Chains (on-chain investigation) (Sep 2026)
13. a16z crypto / ForkLog, Bloomberg's AI agent volumes overestimated fifteenfold (Mar 2026)
14. Visa and Artemis via The Defiant, Visa's Sheffield pegs adjusted x402 volume at $19M (Jul 2026)
15. Crypto Briefing, Base surpasses 20M agentic payment transfers in 90 days (Jul 2026)
16. Coinbase via Bitcoin World / CryptoRank, AI agents 205 million transactions (Aug 2026)
17. Crypto Briefing, Solana commands ~70% market share in x402 activity (Sep 2026)
18. Four Pillars, x402 on Polygon: Real or Gamed? (Feb 2026)
19. Allium, base.agents.x402_metrics_daily documentation (2026)
20. XRPL AI Hub via KuCoin, XRP Ledger processes 8M+ x402 payments in 30 days (Oct 2026)
21. Visa, Visa and partners complete secure AI transactions (Dec 2025)
22. 247 Wall St, Ripple is now a premier member of the x402 Foundation (Jul 2026)
23. Flagship Advisory Partners, Market Tracker: Agentic Commerce, August 2026 (Aug 2026)
24. Alipay Ant Group, Alipay AI Payment exceeds 120 million transactions in one week (Feb 2026)
25. Cloudflare, Introducing AI Crawl Control (Cloudflare blog) (Aug 2025)
26. Cloudflare Radar via Leadgen Economy, Cloudflare Radar 2025 Year in Review (May 2026)
27. 8004scan via Hashlock, Why AI agents need on-chain identity (May 2026)
28. Linux Foundation, Operational launch of the x402 Foundation (Jul 2026)
29. MediaPost, Google agentic shopping launches with 60 partners (Sep 2025)
30. InfoQ, Google UCP (Jan 2026)
31. Shopifreaks, OpenAI launch of Instant Checkout (Oct 2025)
32. Stripe, Everything we announced at Sessions 2026 (Apr 2026)
33. ForkLog, OpenAI abandons Instant Checkout feature in ChatGPT (Mar 2026)
34. Cryptonews via Yahoo Finance, Stablecoin supply hits $315B in Q1 (Apr 2026)
35. Coinbase, Q4 2025 shareholder letter (Feb 2026)
36. McKinsey and Artemis, Stablecoins in payments: What the raw transaction numbers miss (Feb 2026)
37. McKinsey, The agentic commerce opportunity (Oct 2025)
38. Morgan Stanley, Agentic commerce research (via Digital Commerce 360) (Dec 2025)
39. Juniper Research via TechNode, Agentic commerce spending can reach $1.5 trillion by 2030 (Apr 2026)
40. Gartner, Top strategic predictions for 2026 and beyond (via Digital Commerce 360) (Nov 2025)
41. ACI Worldwide / YouGov, Only 7% of fashion shoppers trust AI to buy for them (Sep 2026)
42. Riskified, Agentic Commerce Pulse Q1 2026 (via HomePage News) (Apr 2026)
43. Braze, 2026 Braze Customer Engagement Review (Feb 2026)
44. Checkout.com, Agentic Commerce 2026 (2026)
45. Checkout.com via Forkast, 89% of merchants prepare for agentic commerce, only 3% see agent transactions (Jul 2026)
46. Mastercard, New trust and intelligence services (Sep 2026)
47. Congressional Research Service, Stablecoin Legislation: An Overview of the GENIUS Act of 2025 (P.L. 119-27) (Aug 2026)
48. Crowdfund Insider, GENIUS Act deadline passes without final stablecoin rules (Jul 2026)
49. AIMA, FCA publishes final rules for the new UK cryptoasset regime (Jun 2026)
50. Ant International, Ant International launches open-sourced Agentic Mobile Protocol to drive AI commerce (Apr 2026)
51. Ant International via The Asian Banker, Ant International expands Agentic Mobile Protocol rollout, begins KYA interoperability framework collaboration with Mastercard and Visa (Sep 2026)
52. PYMNTS, Visa introduces Trusted Agent Protocol for AI shopping (Oct 2025)
53. Mastercard, Verifiable Intent (Mar 2026)
54. Monetary Authority of Singapore, Safeguards for Agentic Finance at Runtime (SAFR), version 1.1 (Oct 2026)
55. Global Payments, Consumers expect AI to make 15% of their purchases within five years (Sep 2026)
56. Alipay+ and S&P Global via Newswire, Interoperability and trust key to AI commerce (Sep 2026)
57. TRM Labs, Who's actually paying? Measuring AI agent payments onchain (Sep 2026)
58. HUMAN Security via PPC Land, AI agent traffic dips in May, but blocking rates keep climbing (Jun 2026)
59. UCP Checker, UCP census statistics, read 19 Oct 2026 and confirmed to the author by Benji Fisher (19 Oct 2026)