Something happened last month that I keep thinking about. A friend of mine asked his AI assistant to buy a new ergonomic keyboard. He gave it a budget, said he wanted something with split layout and mechanical switches, and told it to go ahead and order. The AI researched options, compared reviews, checked availability at three retailers, found the best price including shipping, and completed the purchase. My friend never visited a single product page. He never read a single review. He never even saw the keyboard until it arrived at his door.
This is agentic commerce, and it's not a hypothetical future scenario. It's happening now, in March 2026, at a scale that should concern anyone who makes money from web traffic.
The agentic AI market is forecast to exceed $30 billion by 2028. Traffic from AI sources has jumped 1,200 percent for retailers. And according to one recent survey, 70 percent of consumers say they're at least somewhat comfortable with an AI agent making purchases on their behalf. We went from "nobody would let a robot buy things for them" to "most people are fine with it" in roughly eighteen months. That's a faster consumer behavior shift than the adoption of mobile shopping, and it's going to reshape how online commerce works at a fundamental level.
What agentic search actually means
I should define what I'm talking about because the term "agentic" gets thrown around loosely these days.
An AI agent, in the commerce context, is a system that can independently research, evaluate, and execute a purchase on behalf of a user. The user sets parameters — budget, preferences, constraints — and the agent handles everything else. It queries APIs, scrapes product pages, evaluates prices and availability in real time, reads reviews, compares options, applies discount codes, processes payment, and confirms the order. The human approves the final decision (usually), but they don't click through each step manually.
This is different from AI-assisted search, where the AI helps you find information but you still make the purchase yourself. It's different from recommendation engines, which suggest products but leave the browsing and buying to you. Agentic commerce removes the human from most of the transaction funnel. The AI doesn't just find the product. It buys the product.
Analysts project that by 2030, 20 to 30 percent of all online transactions will involve AI agent mediation at some stage of the funnel. That projection might be conservative given the current adoption curve. Five major shopping agent platforms launched between October 2025 and February 2026 alone.
Why your website might never see the customer
Here's the part that keeps me up at night as someone who works in SEO.
When a human shops online, they visit your website. They browse. They read descriptions. They look at photos. They check reviews. Each of those interactions is an opportunity for your brand to make an impression, to build trust, to upsell, to capture an email address. Your website is a conversion machine because humans interact with it directly.
AI agents don't browse your website the way humans do. They query your product data through APIs, structured feeds, or by parsing your pages programmatically. They don't see your beautiful hero image or your carefully crafted brand story. They see product attributes: price, availability, specifications, reviews, shipping time. Your website's design, user experience, and brand personality become largely irrelevant when the customer never sees them.
This isn't theoretical. Shopping agents already bypass traditional product pages entirely when a retailer offers API access or structured product feeds. The agent pulls the data it needs, compares it against alternatives, and either adds the item to cart or moves on. The entire interaction might take two seconds. Your landing page optimization, your A/B tested headlines, your social proof widgets — none of it matters if the purchasing decision is made by a machine parsing your product schema.
I realize this sounds bleak. It's supposed to. The sooner we acknowledge how fundamentally this changes things, the sooner we can start adapting.
Machine readability is the new user experience
If AI agents are making purchasing decisions based on structured data rather than visual experiences, then optimizing your structured data is no longer a nice-to-have SEO task. It's your primary competitive lever.
Schema.org markup has always been recommended practice. But in a world where AI agents are your customers, schema markup is how you communicate. It's your storefront for machines. Product schema with accurate pricing, availability, condition, and review data determines whether an AI agent can even consider your product. Missing or incomplete schema means the agent literally cannot see you.
Think about what that means. For the past two decades, SEO has been about making your content visible and attractive to humans who find it through search engines. Agentic commerce adds a parallel requirement: making your content visible and evaluable to machines that find it through APIs and structured data. The strongest players in this new landscape will be the ones whose product data is accurate, whose inventory signals are real-time, whose reviews are enriched with structured markup, and whose checkout flows can accept agent-led transactions.
I want to spend some time on that last point because it's one most people haven't thought about yet. AI agents need to be able to complete the checkout process programmatically. If your checkout requires solving a CAPTCHA, that blocks agents. If it requires creating an account with email verification, that slows agents down. If your payment flow can't be initiated via API, agents will buy from your competitor instead. The sites that make it easiest for AI agents to complete transactions will capture a disproportionate share of agentic commerce.
The information co-location strategy
Beyond structured data, there's a content strategy question: how do you ensure AI agents have all the information they need to recommend your product?
I call this "information co-location," and the principle is simple. AI agents make decisions based on the information they can access at the point of evaluation. If the agent is comparing three keyboards and yours is the only one with complete specifications, verified reviews, and clear compatibility information all available in one structured location, you win. Not because your keyboard is better, but because the agent has enough information to confidently recommend it.
Think about how differently this works from traditional SEO. In traditional search, you might spread information across multiple pages — a product page, a FAQ page, a compatibility guide, a review page — and link them together. A human visitor would click around and assemble the full picture. An AI agent wants all of that information co-located and structured. It's not going to follow your internal links and synthesize information from five different pages. It's going to evaluate what's immediately available and move on.
This means your product pages need to be significantly more information-dense than they've traditionally been. Specifications, compatibility data, comparison points, use case recommendations, and aggregate review data should all be present on the product page itself, structured in schema markup that an agent can parse instantly. The days of thin product pages that exist mainly as conversion funnels are numbered.
Pricing transparency and competitive positioning
AI agents are ruthless comparison shoppers. They can evaluate pricing across dozens of retailers in milliseconds. If your price isn't competitive, no amount of brand loyalty will save you — because the AI agent has no brand loyalty. It optimizes for the parameters its human set: price, quality, speed, whatever.
This means pricing transparency becomes more important than ever. Hidden fees, complex shipping calculations, and dynamic pricing that changes based on browsing history all work against you in an agentic commerce environment. The AI agent will calculate total cost including everything, and if your transparent total is higher than a competitor's, you lose the sale. You might not even know you were in the running.
There's an interesting counterpoint here, though. Some AI agents are being trained to factor in quality signals and reliability metrics alongside price. An agent that learns your store consistently ships on time, rarely has returns, and maintains accurate inventory data might recommend you even at a slight price premium. Reputation, in an agentic world, gets quantified. And the quantification is based on structured signals — shipping accuracy, return rates, review sentiment — rather than brand perception.
Preparing your tech stack
Let me get practical about what your engineering team should be thinking about.
First, API access. If you're an e-commerce retailer of any meaningful size, you need a product API that exposes your catalog data in a machine-readable format. This isn't just for AI agents — it's also for Google's own shopping systems, price comparison engines, and marketplace integrations. But the agentic commerce wave makes it urgent rather than optional.
Second, real-time inventory signals. AI agents that recommend an out-of-stock product create a terrible user experience. The human gets excited about the recommendation, tries to buy, and finds the item unavailable. They blame the AI, but they also lose trust in your store. Real-time inventory accuracy, communicated through structured data and API endpoints, prevents this.
Third, agent-compatible checkout. This is the hardest technical lift for most retailers. Your checkout flow needs to support programmatic transactions — API-based cart addition, payment processing through standard protocols, and order confirmation that can be relayed back to the agent. Some platforms are building "agent checkout" endpoints specifically for this purpose. If your e-commerce platform doesn't support it yet, start asking your vendor about their roadmap.
Fourth, review enrichment. AI agents weight reviews heavily in their decision-making. But not all reviews are equally useful to an agent. Structured review data — ratings broken down by specific attributes like "build quality" or "customer support" or "ease of setup" — is far more useful to an agent than a blob of unstructured text. Platforms that break reviews into structured attribute ratings will be preferred by agents doing comparison evaluation.
Content strategy for an agentic world
Beyond the technical requirements, there's a content strategy shift that I think most marketers are going to struggle with.
Traditional content marketing assumes a human reader. You write blog posts, buying guides, and comparison articles to attract humans doing research before a purchase. Those humans then convert on your site. The content is designed to be persuasive, engaging, and optimized for human attention patterns.
In an agentic world, much of that research is being done by AI. The AI agent doesn't need to be persuaded. It needs to be informed. It doesn't respond to emotional appeals, urgency tactics, or social proof the way humans do. It responds to data.
This doesn't mean content marketing is dead. It means the purpose of content shifts. Your buying guides and comparison articles still serve the AI agents doing research — but those articles need to be optimized for machine extraction rather than human persuasion. Clear data tables. Unambiguous specifications. Objective comparison criteria. Factual claims with supporting evidence.
Some content still needs to target humans directly. Brand awareness content, thought leadership, and community-building content don't have agentic equivalents. These become more important, not less, in a world where transactional searches are increasingly handled by agents. Your relationship with the human customer is built through the content that agents can't replace.
Where this is heading
I'll be candid about what I don't know. I don't know how fast agentic commerce will actually scale. The 20 to 30 percent of transactions by 2030 projection seems reasonable, but consumer technology adoption is notoriously hard to predict. It could be faster. It could stall if early agent experiences are poor enough to erode consumer trust.
What I do know is that the technical infrastructure for agentic commerce is being built right now, aggressively, by every major technology company. Google, Apple, Amazon, and OpenAI all have shopping agent capabilities in various stages of deployment. The investment is too large and too widespread for this to be a niche phenomenon. Whether it's 20 percent of transactions in 2030 or 10 percent, it's going to be significant enough to reshape online commerce.
The retailers and brands that start optimizing for agentic discovery now will have a meaningful advantage. Not because the current AI agents are that sophisticated, but because the habits and infrastructure you build today compound over time. Getting your structured data right, building API access, and rethinking your content for machine readability — these are investments that pay dividends regardless of exactly how fast the agentic commerce transition happens.
The web was built for humans. For twenty-five years, that was the only audience that mattered. Now there's a second audience, and it has money to spend. Making sure they can find you isn't optional anymore.