Reference🤖 Agentic payments & AIIntermediate⏱ 13 min read

🛒 Agentic commerce: what it means for merchants

Getting found by agents, selling to machines, and measuring traffic that is no longer human: a 2026 playbook for staying on the AI radar

A channel that is already taking off

“Agent” traffic covers visits to an online store that come from an AI source, whether a human follows the link or an agent browses the site on its own. Adobe Analytics tracks this traffic across more than a trillion visits to US retail sites and has measured triple-digit growth since late 2024. The biggest change is in its quality. AI traffic long converted poorly, and it now converts better than conventional traffic.

+393 %
AI traffic to US retail sites in Q1 2026 (vs. Q1 2025)
Adobe Analytics, Apr. 2026
+42 %
conversion vs. non-AI traffic (March 2026), up from -38% in March 2025
Adobe Analytics
+37 %
revenue per visit for AI-referred traffic
Adobe Analytics, 2026
+48 %
time on site for visitors arriving from an AI source
Adobe Analytics, March 2026

The reversal comes down to where the site sits in the purchase journey. When an agent recommends a store, the visitor (or the buying agent) arrives pre-qualified: the need, the budget, and the comparison have already been handled upstream. The site now comes in when the order is confirmed, not when the shopper needs persuading.

ℹ️
What about Europe?
Europe is following a few quarters behind, with Wero (EPI) extending its payments to e-commerce in 2026. Worldline and ING ran Europe’s first live agentic transaction in June 2026, which puts local rails at the early-deployment stage.

Getting found: agentic discoverability

Agentic discoverability is how easily a software agent can read and use a merchant’s catalog. An agent does not browse pages the way a human reader does. It relies on the structured data the site publishes. If a machine cannot read a catalog, it does not show up in the agent’s recommendations, however much the merchant spends on ads.

🗂️
Product feeds
A clean, complete feed (Google Merchant Center-style) is still the foundation, because agents rely heavily on these standardized catalogs.
🏷️
schema.org / JSON-LD
Product markup gives the agent price, stock, and reviews without rendering the page. The minimum set: name, brand, description, image, price with validity date, availability, and aggregate rating.
🔌
MCP and catalog APIs
The Model Context Protocol exposes real-time stock and prices to agents. Shopify has enabled it by default on its agentic storefronts since June 2026.
Minimum agent-readable product markup (JSON-LD)
{
  "@context": "https://schema.org",
  "@type": "Product",
  "name": "Askia trail running shoes, size 43",
  "brand": { "@type": "Brand", "name": "Askia" },
  "image": "https://example.com/img/askia-trail.jpg",
  "aggregateRating": { "@type": "AggregateRating", "ratingValue": 4.6, "reviewCount": 218 },
  "offers": {
    "@type": "Offer",
    "price": 114.90,
    "priceCurrency": "EUR",
    "priceValidUntil": "2026-09-30",
    "availability": "https://schema.org/InStock"
  }
}
🔑
Product data is your new SEO
You cannot “outbid” competitors with ad spend in the agentic channel, so the quality of product attributes (descriptions, sizes, compatibility, reviews, images) is the main lever for visibility. Merchants that fill in more than 95% of their key attributes get recommended by agents far more often.
  • Audit your 100 best-selling products: their JSON-LD markup must be complete and accurate.
  • Your robots.txt file and web application firewall (WAF) must not turn away legitimate agent bots, which can be told apart from hostile ones.
  • To measure visibility, ask ChatGPT, Gemini, or Perplexity to recommend products in your key categories and see whether you appear.

A “machine-readable” checkout

An agentic checkout is an order flow exposed as an API, so an agent can buy without a payment page ever being displayed. Standardized calls between the agent and the merchant replace forms and buttons. The agent submits the order, and the merchant accepts or declines it, collects payment, calculates taxes, and fulfills it through its usual processes.

1 · Discovery & catalogstructured product feedsschema.org / JSON-LDMCP (exposed tools)2 · Agent identityWeb Bot Auth (IETF)Trusted Agent Protocolagent directory3 · Mandate & consentAP2: Intent / Cart / Paymentagentic delegated tokenKYA4 · Checkout & settlementACP (Agentic Commerce Protocol)network tokenx402 / USDCAP2ACPx402UCPTAPnative coveragepartial or delegated coverageAP2, UCP (Google)ACP (OpenAI × Stripe)x402 (Coinbase)TAP (card networks)No protocol covers all four layers: the question isn't “which one to choose” but “which ones to stack.”
RouteHow it worksIntegration effortWho it suits
ACP through the merchant’s PSPStripe, PayPal, and others expose the merchant’s checkout to OpenAI’s protocol; existing Stripe customers can turn on the agentic channel with very few code changesLow to mediumMerchants already with an ACP-compatible PSP
E-commerce platformShopify (agentic storefronts, UCP), Etsy, and similar platforms syndicate the merchant’s catalog to ChatGPT, Gemini, Copilot, or PerplexityLowMerchants on a SaaS platform
Direct integrationImplement AP2/UCP or Visa’s Trusted Agent Protocol (public specs on GitHub) on your own stackHighLarge merchants and marketplaces with in-house engineering teams
Three ways to accept purchases from agents (July 2026)
⚠️
Watch the fees
The agentic channel comes with its own cost. OpenAI, for example, charges a service fee of about 4% on Instant Checkout transactions from Shopify merchants. That fee comes on top of standard acquiring fees, so it belongs in your margin-by-channel calculations now.

The real issue: disintermediation

Disintermediation here means the customer relationship shifts from the merchant to the agent platform that runs the purchase. Who owns that relationship becomes unclear once the agent controls the interface. The merchant then risks becoming an interchangeable supplier behind an API, compared constantly on price alone, with no visible brand, no merchandising, and no upselling.

⚠️
The Amazon v. Perplexity precedent
In November 2025, Amazon sued Perplexity, accusing its Comet agent of posing as a human browser to shop on its platform. In March 2026, a federal judge blocked Comet’s access to Amazon. Legally, then, merchants control who gets into their store. The flip side is that a merchant that blocks all agents indiscriminately cuts itself off from the agentic channel.
🏷️
Algorithmic price wars
Agents that compare everything, all the time, squeeze margins. Differentiation shifts to data: delivery times, returns, warranties, verified reviews.
👤
Losing customer data
The shopper “belongs” to the agent platform, so access to order data has to be negotiated with it. The post-purchase experience (shipping, emails, customer service) remains the merchant’s territory.
🧾
A new platform tax
The agentic channel has its own P&L: agent commissions, PSP fees, and the cost of structured data. It needs to be measured separately from other channels.

The workable stance is neither a total block nor an open door. It is a formal agent access policy that states which agents are allowed in, which part of the catalog they can reach, and on what pricing terms, much as merchants already manage marketplaces.

New KPIs to track

E-commerce KPIs are built on sessions and conversions attributed to human visitors. Agentic traffic calls for its own metrics, which should be separated from human traffic in analytics tools starting in 2026. The table below lists six, with their definitions and uses.

KPIDefinitionWhy it matters
Share of agent trafficSessions identified as AI (AI referrers, Web Bot Auth signatures) ÷ totalMeasures exposure to the channel; growing fast everywhere
Agentic share of GMVRevenue initiated by an agent ÷ total revenueThe real business metric; track it by protocol (ACP, UCP…)
Agent conversion rateOrders ÷ agent sessionsBenchmark: AI traffic now converts better than average (Adobe, March 2026)
Data coverage rateProducts with complete key attributes ÷ catalogDirectly correlated with visibility in agent recommendations
Agent block rateLegitimate agent requests blocked by the WAF ÷ totalA misconfigured WAF can make a merchant invisible in the channel without anyone noticing
Net margin by channelMargin after agent commissions and payment feesThe agentic channel has its own cost structure
A merchant’s agentic commerce dashboard
+138 %
AI traffic to US retail sites, year over year, still in May 2026
Adobe / Digital Commerce 360, June 2026
+12 %
engagement from AI-referred visitors vs. conventional traffic
Adobe Analytics, March 2026

Checklist: getting ready in 2026

The list below ranks the work needed to prepare for the agentic channel by priority. Most online merchants can complete the first six items in a quarter.

  • Audit your product data: complete JSON-LD Product markup (price, stock, reviews, priceValidUntil) across the entire active catalog.
  • Check agent access: WAF and robots.txt rules that tell signed agents (Web Bot Auth) apart from hostile bots, rather than a blanket block.
  • Measure: segment traffic and revenue by AI source in your analytics; build the agentic KPI dashboard.
  • Ask your PSP: Does it support ACP? UCP? The Trusted Agent Protocol? On what timeline? Most major PSPs published an agentic roadmap in 2026.
  • Test a channel: turn on Instant Checkout (if eligible through Stripe, Shopify, or Etsy) or your platform’s catalog syndication for a limited set of products.
  • Calculate channel margin: build agent commissions and fees into the P&L before pushing volume.
  • Set an agent policy: which agents are allowed, for which products, and with what shared data; then document it in legal terms.
  • Train your teams: SEO now also means AEO (agent engine optimization), and customer service needs to know how to handle an order placed by an agent.
✅
The cost of doing nothing
Structured data and agent traffic measurement cost little, and they help traditional SEO too. The main risk in 2026 is not investing too early. It is being invisible when the channel becomes significant.
Merchant-side ecosystemShopifyStripeAdyenPayPalEtsyCLCloudflareVisaMastercard