Agentic Commerce: How to Stay Visible When AI Makes the Buying Decision

"I need hiking boots for a high-alpine trek." Someone is typing that into Google's AI Mode right now. Not into classic search. Not into your shop's filter bar. Into an AI assistant. And the AI doesn't respond with ten blue links – it compares models, weighs waterproofing against weight, checks delivery times, reads across reviews, and finally puts forward three specific boots to choose from. With a reason for each. And because the AI understands the full context, it recommends the right extras straight away: the matching socks, a waterproofing spray. The customer has made their buying decision – accessories included – before ever seeing a single online store. Welcome to agentic commerce, and to the question that will shape your revenue tomorrow: will an AI even consider your brand?

The Traffic Shift: Why Revenue Is Leaving the Search Results Page

There's a stubborn misconception that AI shopping is a niche topic for early adopters. A look at the numbers puts that to rest.

AI referral traffic – visitors reaching shops through AI assistants – has multiplied over the past year, according to Adobe. At the same time, classic search clicks are collapsing: Ahrefs reports sharp drops in organic click-through rates for information-driven queries, and around 60 percent of all Google searches now end without a single click. The keyword here is zero-click search: the answer appears directly in AI Overviews, and the user never has to leave the search page.

This isn't a statistic to file away. It's your revenue source shifting in real time. The traffic isn't disappearing. It's simply relocating – away from the results list and into the AI's answer. And whoever doesn't appear there doesn't exist for the customer.

Retailers used to fight for the number-one spot on Google. Tomorrow they'll fight just to be recommended by an AI at all.

Beyond Checkout: How Agentic Commerce Takes Over the Entire Buying Journey

Whenever agentic commerce comes up, the conversation lands on payment surprisingly fast. "Can the AI buy on its own?" "How secure is AI checkout?" Fair questions – but they miss the bigger picture.

Picture the classic sales funnel: awareness at the top, then interest, advice, comparison, recommendation – and right at the end, as the smallest step, the purchase. That's exactly where the thinking goes wrong. Checkout is the last step. But the AI takes over every step before it first.

Product selection. Advice. Comparison. Recommendation. These are the moments where your category pages, filters, product descriptions, and buying guides used to make the difference. Those moments are now moving into the AI assistant. Once the AI has prepared the decision, checkout is merely the execution of an outcome that's already been settled. So your revenue isn't decided at the register – it's decided far higher up the funnel, in a place where you may no longer be visible at all.

Definition: What Agentic Commerce Actually Means

Now that it's clear *where* the shift happens, a clean definition is worth having.

Agentic commerce describes shopping processes in which AI agents independently research, compare, advise, and in some cases buy – on the user's behalf.

The person sets a goal ("the right hiking boots for the Alps"), and the agent handles the rest. These agents come in two basic forms: as your own shop agent, deployed directly on your platform to advise customers – and as a cross-platform assistant like ChatGPT, Google Gemini, Perplexity, or Google AI Mode, which researches across thousands of providers. The distinction matters: with your own agent, you're playing at home. With a cross-platform assistant, you first have to be found at all.

The Three Maturity Levels of Agentic Commerce

Not every AI purchase is equally advanced. It helps to distinguish three stages:

1. Assisted – The AI helps research and compare, but the human clicks "Buy" themselves.

2. Delegated – The AI proposes specific products and prepares the purchase in full; the human only confirms.

3. Autonomous – The AI buys independently within defined limits, for example reorders or routine purchases.

Most customers today move between stages 1 and 2. But even that is enough to redistribute your visibility entirely.

Market Check: The Shift Is Moving Faster Than Many Assume

While many companies are still debating whether agentic commerce is "even relevant yet," the major platform providers are already creating facts. Not a future scenario – live operation.

OpenAI led the way in September 2025: with Instant Checkout and the Agentic Commerce Protocol (ACP), purchases were meant to happen right inside ChatGPT. Just six months later, in March 2026, OpenAI reversed course and scrapped in-chat checkout – conversions were far weaker than sending shoppers to the retailer's own site. The new stance is telling: ChatGPT advises, compares, and guides all the way to the cart, but hands the actual purchase back to the shop. Advice yes, payment no.

Google is pursuing the opposite strategy – and thinking it far bigger. With Universal Cart, unveiled at I/O 2026, Google wants to accompany the customer across its entire ecosystem: whatever you discover in Search, in a conversation with Gemini, on YouTube, or even in Gmail lands in a single, cross-device cart. In the background, Gemini tracks price histories and flags price drops and restocks. Payment runs directly through Google Pay – in a few taps, without ever leaving the Google universe. Technically, it all rests on the Universal Commerce Protocol (UCP) as a shared agent language and the AP2 protocol (Agent Payments Protocol) for secured payments. Google doesn't want to handle individual purchases – it wants to own the entire shopping journey.

Perplexity takes a third route: the assistant researches and recommends, but leans entirely on a deep PayPal integration for the purchase. Perplexity operates no payment system of its own – the transaction, payment included, is handled by PayPal. So whoever appears in "Buy with Pro" is selling through someone else's payment layer.

Amazon comes at it from another angle. Its Rufus assistant became "Alexa for Shopping" in 2026, and the goal isn't the spectacular auto-purchase but the fight against choice overload. Features like "Help Me Decide" compare products using reviews and listing content and put forward a reasoned recommendation. Add to that easier reorders and smart cross- and upselling – fed by what Amazon knows about its customers. Less decision fatigue, more relevant add-on purchases.

Walmart, in turn, thinks about agentic commerce as omnichannel: its assistant Sparky doesn't just interpret prompts like "plan me a camping trip for four" and return curated recommendations instead of a results list – it's meant to actively advise customers online and in-store, suggest add-on products, and guide them through the range. An advisor instead of a search box – online and offline.

However different the paths, the direction is clear: advice, recommendation, and payment are all moving into the AI. And that's exactly where the crucial gap opens up. A Salesforce study finds that 72 percent of companies aren't ready for it. That's where the real story lies.

AI Readiness: The Real Challenge Is Your Data

Here comes what may be the most important message of this article, and it's pleasantly unspectacular: You don't necessarily need your own AI agent. You first need clean data.

The temptation is strong to treat agentic commerce as a technology project – "let's build a chatbot." But an assistant drawing on unstructured, incomplete, or contradictory product data gives poor recommendations. And an external AI that can't read your data cleanly won't recommend you at all.

AI readiness means your product information is so structured, complete, and machine-readable that an AI can understand it, classify it, and recommend it. That's the foundation everything else stands on. Without it, any agent strategy runs into the void. So the maturity level that comes first isn't "autonomous agent" – it's foundation: AI readiness.

The Price Trap: Without Good Data, AI Decides on Price Alone

This is the sentence worth remembering:

Without soft data, AI decides on nothing but price, availability, and delivery time.

Because what does an AI do when it compares two hiking boots and finds nothing beyond bare technical specs? It falls back on the one thing that's unambiguously machine-readable: the hardest number. Price. And now you're in a pure price war – the one almost no one wants to win.

What an AI needs instead to recommend you *beyond price* are the soft, context-rich signals: ratings and reviews that prove quality and fit; community and user-generated content that show how a product performs in real life; rich product data on materials, use cases, fit, and compatibility; demonstrable delivery quality; and structured data that makes all of this unambiguously readable for machines. These soft signals are your new competitive advantage – the reason an AI recommends your product even when it isn't the cheapest.

The Agentic-Ready Checklist: 5 Building Blocks for Your AI Visibility

What does AI readiness look like in practice? Five building blocks – and, for each, why the AI needs exactly that.

1. Structured product data. An AI doesn't read pretty prose, it reads fields. When material, size, use case, and attributes sit cleanly in defined fields, the AI can match your product precisely to a customer's request. Leave them out, and you're guessing – and so is the AI.

2. Schema markup (Schema.org). The shared language in which you tell machines what's on your page: this is a product, this the price, this the rating, this the availability. It's the difference between "possibly a hiking boot" and "hiking boot, waterproof, size 43, €189, 4.6 stars, in stock now."

3. APIs and interfaces. Agents don't want to scrape, they want to query. Through open interfaces – and, increasingly, protocols like ACP, UCP, and AP2 – an AI can pull prices, stock, and availability in real time and trigger the purchase. Without an interface, your range stays a black box.

4. Logistics and delivery quality. Availability and delivery time are hard decision criteria for agents. Whoever delivers reliably – and shows it cleanly in the data – gets recommended preferentially, because the AI wants to keep the customer promise.

5. Trust signals. Ratings, verified reviews, certifications, return rates, brand reputation: these are the trust signals an AI uses to decide whether it can recommend you in good conscience. Trust has become machine-readable – if you make it readable.

Sounds like a lot? It isn't. None of this is science fiction or cutting-edge tech – it's clean data work. The real edge doesn't come from it being complicated; it comes from the fact that most players simply haven't done it yet.

Brand, Reimagined: It Doesn't Disappear – It Changes Its Location

This is the point where many start to feel uneasy. If the AI compares and recommends – what's left of my brand? The answer: your brand doesn't disappear. It changes its location.

Customers used to find your shop. They typed your name into Google, landed on your homepage, experienced your design, your language, your promise. Today the AI finds your data. It's no longer the homepage that decides, but the quality and structure of the information you provide. Tomorrow the AI decides whether your brand becomes part of the buying decision at all – whether you show up in the set of two or three recommended options, or don't appear in the first place.

That's no reason to panic, but it is a reason to act. Brand doesn't become less important in agentic commerce. It simply moves from the surface into the substance – from the beautiful landing page into the solid data behind it.

Frequently Asked Questions About Agentic Commerce (FAQ)

What is agentic commerce?

Agentic commerce refers to shopping processes in which AI agents independently research, compare, advise, and in some cases buy – on the user's behalf. The person sets a goal, and the AI agent takes over product selection, comparison, and recommendation, often before the customer has seen an online store at all.

How does agentic commerce work?

A user describes their need in natural language to an AI assistant like ChatGPT, Gemini, or Perplexity. The agent searches product data across many providers, weighs criteria such as suitability, reviews, price, and delivery time, and puts forward a reasoned recommendation. Through protocols like ACP, UCP, or AP2 and open APIs, the purchase can in some cases be completed right in the chat.

How do I prepare my shop for agentic commerce?

The first step is AI readiness: structured, complete, machine-readable product data. In concrete terms, that means structured product attributes, Schema.org markup, open APIs/interfaces, clean logistics and availability data, and trust signals such as reviews and certifications. Only on that foundation do your own shop agents or deeper AI features pay off.

How does AI find products?

AI agents find products through structured data and interfaces – not through pretty layouts. What matters are machine-readable attributes, schema markup, current availability and price data, and trust-building signals. When context-rich "soft" data is missing, the AI narrows its choice down to price, availability, and delivery time.

Glossary: The Key Terms

AI readiness – The state in which product information is so structured, complete, and machine-readable that AI agents can understand, classify, and recommend it. The foundation of any agentic commerce strategy.

Schema.org – An open standard for structured data that lets websites tell machines unambiguously what their content means (e.g. "product," "price," "rating," "availability").

Agent Commerce Protocol (ACP) – A protocol driven by OpenAI that defines how AI agents, shops, and payment providers communicate to complete purchases directly in the assistant. Related approaches: UCP (Universal Commerce Protocol) and Google's AP2.

Trust signals – Trust-building, machine-readable signals such as ratings, verified reviews, certifications, return rates, and brand reputation. They help determine whether an AI recommends a product.

Zero-click search – Search queries that end without a click on a result because the answer appears directly in the search interface (e.g. in AI Overviews). A central driver of the traffic shift.

Conclusion

Agentic commerce widens the playing field: a well-designed, compelling shop still matters – but behind it, you'll increasingly need clean data that an AI can read, classify, and recommend. One doesn't replace the other; it's added to it. The order is clear: AI readiness first – structured data, schema markup, APIs, delivery quality, trust signals – then your own agents and features. Whoever starts getting their data foundation in order now secures a spot in tomorrow's recommendations. Whoever waits leaves the decision to price.

Want to know how AI-ready your product data really is? Get in touch and we'll run a structured AI readiness check for your shop. Talk to us.