For most of the internet's commercial history, sellers controlled the funnel. You ran ads, the click landed on your site, your design and copy persuaded a human, your checkout collected the money, and your CRM owned the relationship afterwards. Search engines and recommendation systems shaped what humans noticed, but the human did the buying.

Agentic commerce changes who is in the room. Increasingly, an AI agent sits between the buyer and the open web. The buyer says "find me a decent espresso machine under €600 that ships to Estonia and has good reviews from people who actually drink espresso every day." The agent does the searching, the comparing, the question-asking and — sometimes — the paying. The seller's website is now one input among many, read and judged by software, not by the buyer.

That is the shift worth understanding. Not "AI can do checkout." But where the moments of intent, payment authority, trust and discovery actually live now — and what that does to the businesses on the other end.

The funnel is inverting

Three things move at once when commerce becomes agentic. They are easy to miss individually because each one looks small. Together they change the shape of how customers reach a business.

1. Buyer intent is no longer visible to the seller

The old funnel made human intent legible. Someone typed "espresso machine reviews" into Google, clicked your article, scrolled, came back two days later, added to cart. You saw all of that. Analytics, retargeting, email flows — the whole industry was built around watching what humans did once they arrived.

In agentic commerce the human's intent is captured by their agent first. The agent translates a vague brief into structured criteria, holds preferences ("I drink espresso, not lungo"), holds memory ("last time I bought from this brand the delivery was late"), and only then decides which sellers to query. By the time a request hits your store, most of the deliberation has already happened — and the seller has zero visibility into it.

2. Payment authority travels with the task

The second shift is in money. In a normal ecommerce flow, payment is tied to a logged-in human at the checkout page. In agentic flows, payment authority is delegated to the agent ahead of time and travels with the task. The agent does not need to handle the buyer's raw card details. It needs an authorisation that says: this agent, for this buyer, can spend up to this amount, on this kind of purchase, until this date.

Stripe's Agentic Commerce Suite is one of the early infrastructure pieces here. It introduces things like Shared Payment Tokens — a way for an AI agent to pay a merchant without exposing the buyer's credentials. Stripe Link is being extended so that AI agents can transact through it on behalf of a user, with approval flows in between. The detail to notice is not the specific product. It is the design intent: payment is becoming something an agent can carry, scoped and revocable, instead of a one-shot human-at-the-keyboard event.

3. The competitive question becomes whether agents can use you

Old SEO answered: can a human find this business? Agentic commerce asks something more direct: can an agent call this business, understand what it sells, compare it to alternatives, and complete a purchase without a human in the loop?

That is a much harder bar. It implies your products and services exist as data that an outside system can read. It implies your prices, terms, availability and policies are explicit, not buried inside marketing copy or a PDF. It implies your business has APIs, feeds or protocol adapters that an agent can actually call. And it implies your trust signals — reviews, certifications, real customer history — are represented somewhere the agent can verify.

The seller website becomes less central, not less important

A common misread is "if agents do everything, the website does not matter anymore." That gets it backwards. The website, or whatever machine-readable surface a business presents, becomes more important — but its job changes.

The website used to be the place where a human was persuaded. In agentic commerce it becomes the place where the commercial reality of the business is made explicit. That means:

  • A complete, structured product or service catalogue, not just images and prose.
  • Honest, machine-readable prices, including discounts, bundles, taxes and shipping.
  • Policies an agent can quote: returns, warranty, support, delivery windows, service levels.
  • Stock and availability that reflect reality — in real time when it matters.
  • Feeds, APIs or protocol endpoints (such as ACP) so an agent can transact without scraping.
  • Payment readiness: support for tokenised, agent-friendly payment flows like Shared Payment Tokens or wallet-based agent payments.
  • Trust signals an agent can read: real reviews, real customer references, certifications, identity, history.

None of that is glamorous. All of it is the unsexy plumbing that decides whether an agent will recommend you, route a buyer to you, or quietly skip you in favour of a competitor that filled in the blanks.

Internal link

Where this hits your site directly

The companion service to this article is our AEO and AI visibility work — auditing how a brand, product catalogue and policy pages are read by ChatGPT, Gemini, Perplexity and Google AI Overviews, and fixing what is missing.

Open the AEO service →

Agentic visibility is not just SEO with a new name

It is tempting to flatten all of this into "SEO for AI" and call it a day. That misses what is actually changing. Classic SEO is mostly about ranking — being discovered. Agentic visibility is about being understood and used.

Imagine the brief "find me an authentic Italian coffee brand that ships to Tallinn and sources beans transparently." A human Googles a keyword. An agent does something different. It builds a small purchasing specification — origin, certification, freshness, delivery, price band, preference history — and goes looking for businesses that fit. Keyword density on a category page does not solve that. Whether your site clearly states origin, sourcing terms, freshness, delivery, return policy and price in a way the agent can parse does.

The same shift applies to comparison. Agents will compare you against three alternatives a buyer never asked about, on dimensions a buyer never explicitly stated. If the agent cannot extract those dimensions from your pages, it will use the alternatives that filled them in.

Stripe and the infrastructure layer

One reason agentic commerce stopped being a thought experiment in 2025–2026 is that the payment and identity infrastructure caught up. Stripe is the most visible example.

Stripe and OpenAI co-developed the Agentic Commerce Protocol as an open standard for programmatic commerce flows between buyers, AI agents and businesses. The intent is clear: build the merchant integration once and be compatible with multiple AI agents, while the merchant retains control over pricing, fulfilment and the customer relationship. Instant Checkout inside ChatGPT — at launch, US ChatGPT users buying from Etsy merchants — is one of the first concrete uses of the protocol.

The Agentic Commerce Suite is the broader product around it: tools to help businesses sell through multiple AI agents, including Shared Payment Tokens that let an agent transact on a buyer's behalf without holding raw card data. Stripe Link, the consumer wallet, is being extended so AI agents can spend through it with approval flows and, in the future, things like spending limits. None of these features alone is the story. Together they form a payments substrate that assumes the buyer is not at the keyboard and that the agent needs scoped, auditable authority to act.

Instant checkout is not the whole future

It is worth being precise about what in-chat checkout actually is and is not. It is the most visible piece of agentic commerce, and the easiest to demo. It is also not the whole game — and not every merchant has loved it so far.

Walmart, an early participant in ChatGPT in-chat purchasing, reportedly saw conversion roughly one-third of click-out and pulled back. The point is not that agentic commerce is doomed. It is that where the agent should hand back to the merchant is still being figured out. Sometimes a frictionless in-chat purchase wins. Sometimes the buyer wants to see your site, your support, your full options before they commit, and the agent's job is to deliver them there confidently — not to buy on autopilot.

Discovery and handoff matter as much as checkout. Maybe more.

Trust moves into buyer memory

Brand used to live in a human's head. In agentic commerce, brand increasingly lives in the buyer's agent's memory. That memory is durable, structured and quietly compounding. "This brand delivered on time twice." "This shop refused a return that should have been simple." "This vendor's prices match the policy on the website." Over many small interactions, the agent builds a model of who is reliable for which kind of purchase.

That memory is the new flywheel. A business that consistently does what its policies say, ships when it promises, answers what it claims it answers, and is callable through stable interfaces will quietly win share inside agent memory long before anyone calls it a winner. The opposite is also true.

What this means for businesses now

None of this requires a sci-fi rebuild. It requires being honest about the parts of the business that are currently tacit and making them explicit. The questions below are the ones we use with clients when we audit how an agent will read them.

Agentic commerce readiness checklist

  • Can an AI agent understand what you sell, in plain structured form?
  • Can it compare you against a competitor on price, terms, delivery and policy?
  • Are your prices, taxes, shipping and discounts machine-readable, not buried in marketing copy?
  • Are returns, warranty, support hours and service levels written in a way an agent can quote?
  • Is your stock or availability visible somewhere an agent can check?
  • Can an agent call your business through a feed, an API or an agentic protocol like ACP?
  • Are you ready for agent-friendly payments — Shared Payment Tokens, wallet-based agent flows, scoped authorisations?
  • Are your trust signals — real reviews, real cases, certifications — represented in a form an agent can read and verify?
  • Is your brand showing up in the buyer's agent memory through consistent delivery, not just advertising?
  • Do you measure how AI tools currently describe and recommend you, and do you have a way to fix what is wrong?

How RandomForest works on this

Two of our service pages map directly to the work agentic commerce actually requires. Our AEO and AI visibility service audits how a brand is read by AI agents and rewrites the pages, FAQs, schema and signals that decide whether an agent can understand and recommend you. Our Website AI team keeps the content side of that machine-readable surface alive: blog posts, translations, landing pages, policy pages and the supporting copy that AI agents read on the way to a decision.

For WooCommerce stores specifically — where most of the catalogue, pricing and inventory work happens — our AI digital team for WooCommerce wraps an agent setup around the store: product copy, FAQs, campaigns, analytics, AEO checks and the day-to-day store changes you would normally need a developer for.

FAQ

What exactly is agentic commerce?

Agentic commerce is commerce where an AI agent acting on behalf of a buyer performs the work of researching, comparing, deciding and paying. The shift is not just that an agent can buy something — it is that buyer intent, payment authority, trust and discovery move into the agent before any seller's website sees the visitor.

How is it different from regular ecommerce?

In regular ecommerce the seller controls the funnel: site, product pages, ads, checkout, recommendations. In agentic commerce the buyer's agent owns most of that journey. The seller's job becomes being machine-readable, callable and trustworthy enough for an agent to choose, recommend and complete a purchase from the business.

Is agentic commerce only about checkout inside ChatGPT?

No. In-chat checkout is one visible piece. The bigger change is in discovery, comparison, payment authorisation, trust signals and post-purchase service — all of which can happen without a human ever touching the seller's website. Some early in-chat checkout flows have shown lower conversion than click-out, so the design question is where the agent should hand back to the merchant.

What is the Agentic Commerce Protocol (ACP)?

ACP is an open standard, co-developed by Stripe and OpenAI, for programmatic commerce flows between buyers, AI agents and businesses. It lets a merchant build the integration once and be compatible with multiple AI agents, while keeping control over fulfilment, pricing and the customer relationship.

What does this mean for SEO?

Classic SEO optimised for human readers and rankings. Agentic visibility, sometimes called AEO (answer engine optimization), optimises for AI agents that read your site, parse it, compare you to competitors and decide whether to recommend or transact with you. Rankings still matter; being machine-readable and callable matters more.

What should a business do first?

Make the commercial reality of the business explicit and machine-readable: catalogue, prices, policies, delivery and return terms, support hours, service levels and trust signals. Then make the business callable through clean feeds, APIs or agentic protocols. Then track how you appear inside AI agents and where comparison loses you to a competitor.

Sources