For twenty years the path to a sale online was simple to picture: a shopper typed something into a search box, scanned a page of links, clicked through to a store, and bought. Every part of digital commerce, from SEO to ads to product pages, was built around that human in the loop. That human is now getting an assistant. Increasingly, the assistant shops instead of them.
That changes the question every merchant has to answer. It is no longer only "can a customer find my store?" It is "can the AI find my store, understand my products, trust my brand, and transact with me?" If the answer is no, you are not losing a ranking. You are absent from the recommendation entirely, because the AI returns one confident answer, not a page of options to scroll past.
What "agentic" actually adds
We have had AI in shopping for a while: recommendation engines, chatbots, search that understands plain language. What is new is autonomy. An agentic system can carry a goal through multiple steps without handing control back at each one. Ask it for "a waterproof jacket for hiking in the Lake District, under 150 pounds, in stock now," and a capable agent can interpret the need, gather candidates, read specs and reviews, narrow the list, and in a growing number of cases place the order.
For the shopper that is convenience. For the merchant it is a new gatekeeper sitting between you and the customer, one that reads data rather than admiring your design, and that weighs you against every rival in the same instant. The merchants who win are the ones the AI can find, understand, and transact with on day one, without surrendering their brand in the process. That single idea is the spine of this whole series.
The three layers every merchant must own
It helps to break "getting sold by AI" into three distinct layers. They stack, and a gap in any one of them quietly removes you from the sale. Most of this series is about closing those gaps one at a time.
Discovery: can the AI find and understand you?
Before an agent can recommend you, it has to know you exist and grasp what you sell. That means your catalog has to be reachable by AI crawlers, written so it can be extracted, and described in clean, machine-readable data. This is the layer where being absent is most common and most invisible.
The discipline here is GEO, Generative Engine Optimization. If it is new to you, our 7 pillars of GEO guide is the full framework, and GEO Explained shows why AI struggles to see most stores today.
Transaction: can the AI actually buy from you?
Being recommended is not the same as being purchasable. New commerce protocols now let an agent move from "this looks good" to a completed checkout: choosing a variant, confirming availability and price, and placing the order through your store. If the agent hits a wall at the cart, the recommendation converts to nothing.
This is where live tools matter. Beyond static feeds, a store can expose actions an agent calls directly, the idea behind WebMCP and agentic commerce.
Payment: can the agent pay, and do you stay in control?
The final step is money changing hands safely. A new generation of payment rails lets an agent authorize a purchase for a single merchant and basket without ever handling raw card details, and ideally keeps you as the merchant of record, owning the transaction and the customer relationship rather than renting it from a platform.
Control is the theme. The whole point of getting these layers right yourself is to be sold by AI on your terms, not commoditized into a price-and-spec line item in someone else's aisle.
Why now, and not later
The honest case for acting now is not that agentic commerce is already huge. It is that the behavior is growing fast from a small base, which is the best possible moment to get ready. The shift is visible in the data:
- Adoption is climbing steeply. Adobe Analytics measured AI-referral traffic to US retail sites rising roughly 1,300 percent over the 2024 holiday season, and reported it climbing again, up about 693 percent, the following holiday period.
- The forecasts are large. McKinsey projects agentic commerce reaching around $1 trillion in the US and $3 to $5 trillion globally by 2030. Gartner expects roughly 90 percent of B2B purchases to be AI-intermediated by 2028.
- Shopper habits are already moving. Capgemini's research found 58 percent of consumers have replaced traditional search engines with generative-AI tools for product and service recommendations, up from 25 percent a year earlier.
- But the base is still small. Bain notes AI can be up to a quarter of referral traffic for some retailers, yet remains under 1 percent of total retail traffic overall. That gap is the opportunity.
What this series covers
This is the first of eight posts. Each one takes a single slice of the same question, how a merchant gets discovered and sold by AI agents without losing control of the brand, and ends with one concrete thing to do. Here is the road ahead:
- The new shopping aisle: a field guide to where AI buyers actually appear, from ChatGPT and Gemini to Perplexity and Amazon, and the payment rails behind them.
- From SEO to GEO: why getting picked by AI is a different game than ranking on Google, and what genuinely moves the needle.
- Agent-readable by design: feeds, structured data, and the new product page that AI can actually parse.
- Let an agent check out: the commerce and payment protocols, and how to stay the merchant of record.
- MCP and WebMCP: handing agents live tools to search, check stock, and add to cart.
- The disintermediation risk: how to avoid becoming a drop-shipper in someone else's aisle, the honest counterweight to all the optimism.
- Your 90-day playbook: the ordered action list that pulls the whole series into a plan you can start this week.
Do this now: the five-minute audit
Before you change anything, find out where you stand. Open ChatGPT, Gemini, or Perplexity and ask it the way a real customer would, in plain language, to recommend products in your category. Use a specific, natural query, for example "best running shoes for flat feet under $120" or "good gift for a coffee lover under 50 pounds."
- Do you appear at all? If not, you likely have a discovery problem at layer one.
- How are you described? Check whether the AI gets your name, price, and key attributes right, or repeats something out of date.
- Who shows up instead? The competitors the AI names are the ones currently winning the recommendation. That is your real benchmark.
Write down what you see. It is the baseline you will measure everything else against, and it makes the rest of this series concrete instead of theoretical. The AI front door is open. The only question left is whether it can find your store.
Frequently asked questions
What is agentic commerce?
Agentic commerce is shopping carried out by an AI agent on a person's behalf. Instead of only recommending products, the agent researches options, compares them against the shopper's stated needs, and can complete the purchase end to end. For a merchant it means a new kind of buyer: software that reads your catalog data, weighs you against competitors, and decides whether to put you in the basket.
How is agentic commerce different from regular online shopping or SEO?
In traditional online shopping a human browses your site and clicks buy. In agentic commerce an AI assistant sits between the shopper and your store, often answering with one synthesized recommendation rather than ten links. Winning is less about ranking a page a person clicks and more about being the option the AI can read, trust, and transact with. That discipline is called Generative Engine Optimization (GEO).
Is it too early for merchants to care about AI shopping agents?
No, and that is exactly the point. AI-referral traffic to US retail sites jumped roughly 1,300 percent over the 2024 holiday season (Adobe Analytics), yet AI still drives under 1 percent of total retail traffic today (Bain). The behavior is growing fast from a small base, which means the merchants who get readable and transactable now build their advantage before the volume arrives, not after.
What is the one thing a merchant should do first?
Run the audit yourself. Open ChatGPT, Gemini, or Perplexity and ask it to recommend products in your category, the way a customer would (for example, best running shoes for flat feet under $120). Note whether your brand appears, how it is described, and which competitors show up. That five-minute test tells you where you stand at the AI front door before you spend a dollar fixing anything.
Sources & further reading: Adobe Analytics: generative-AI traffic to US retail sites, McKinsey forecast via DigitalCommerce360, Gartner B2B forecast via DigitalCommerce360, Capgemini Research Institute, and Bain & Company: agentic AI in retail. Figures are drawn from the cited reports and forecasts; treat projections as directional.