Network-level data from HUMAN Security shows that 2.31% of agent traffic on e-commerce sites goes to checkout flows. Small in percentage terms. Enormous in kind. They're authenticating, comparing, selecting, and completing purchases.
The instinct, naturally, is to reach for the nearest analogy. Search created SEO, and SEO created an industry. Agencies, tools, consultancies, conferences. A vast intermediary economy organized around one question: how do you get found? The term Answer Engine Optimization appeared as early as 2018. By January 2026, the World Economic Forum was writing about Agentic Engine Optimization. In April, Webflow launched an AEO product in private beta. The intermediary economy is already forming.
At the discovery layer, the analogy holds. Brands cited in AI Overviews earn 35% more organic clicks than uncited brands on the same queries. If you squint, it looks like SEO all over again. Visibility matters. Optimization pays.
But that 2.31% checkout number is the crack in the analogy. SEO optimized for attention. The entire discipline assumed a human who would see a result, click through, browse, and maybe convert. The funnel was leaky by design, and the leakiness created the business model. Every abandoned click was a retargeting opportunity. Every comparison page was an ad surface.
Agents execute. Shopify reports AI-driven orders grew 15x since January 2025. Shoppers arriving from AI services are 38% more likely to buy. The conversion rate goes up because the browsing goes away. A person shopping for a camera might visit five websites. Cloudflare's Matthew Prince has noted that an agent doing the same task will visit 5,000. But those 5,000 visits look nothing like human traffic. They're structured queries, not page views. The ad surfaces along the way don't get seen. The retargeting pixels don't fire in a way that matters.
So what does optimization even look like when the visitor has a job to finish? Can the agent authenticate on behalf of a user? Can it read real-time inventory? Can it complete a transaction through a structured protocol rather than scraping a checkout form? All of them are infrastructure questions.
The uncomfortable corollary for anyone building the next intermediary economy: you cannot optimize what you cannot measure. There is no Search Console for agent traffic. Most analytics platforms can't distinguish an agent visit from a human one. SEO emerged from a tight feedback loop. You changed something, watched the ranking move, learned, iterated. The people already selling AEO tools and services are building an optimization discipline on top of a measurement vacuum. They're offering to tune a system whose responses no one can reliably observe. That's a fundamentally different commercial foundation than the one SEO was built on, and it suggests the intermediary economy that forms here will have a different shape than the one it's borrowing its name from.
Some optimization discipline will likely emerge. But it will probably look like API product management, like building for a customer who happens to be software. The intermediary economy forming around "agent visibility" may find that visibility was only ever the opening act.