Microsoft reported in April that agentic browser traffic on its infrastructure is growing roughly 8,000% year over year. That number is large enough to feel abstract. In practice, it means a growing share of the sessions hitting commercial websites aren't people browsing. They're software completing tasks.
Which is a problem for every platform whose business model depends on influencing human behavior. Advertising, product placement, loyalty programs, recommendation engines, page layout. All of it assumes a person is looking at a screen, making decisions shaped by attention and preference. When the visitor is an agent executing a purchasing task, none of those signals land.
Cloudflare found something quietly revealing about how agents read the web: AI crawlers consume deprecated content at the same rate as current content. The metadata layer that tells humans and search engines "this page is outdated, look over here instead" is invisible to agents that parse everything with equal weight. Canonical tags, freshness signals, the entire architecture of web curation built over two decades. None of it registers. The organizing logic of the human web becomes noise for machine readers, and nobody declared a transition. The old signals just stopped working for a new class of visitor that doesn't know they exist.
AI-attributed orders on Shopify are up 11x since January 2025, with AI-referred traffic growing 7x. But most analytics dashboards show zero conversions from these channels — agents complete transactions via API calls where no JavaScript fires, no cookies set, no thank-you page loads.
Revenue is arriving through a door the measurement system doesn't know exists.
Some companies have started building for this new visitor. Stripe's Machine Payments Protocol, Google's Universal Commerce Protocol, Shopify's agent-facing storefronts. They converge on the same move: replace account creation flows, pricing pages, and subscription tiers with machine-readable interfaces that agents can navigate programmatically. And they all surface the same uncomfortable question. If the storefront, the search bar, and the recommendation carousel are built for human attention, what's left when the customer doesn't have any?
The tension is sharpest at Amazon. Its advertising business generated $68.6 billion in 2025, built on human shoppers seeing sponsored product placements. Amazon is simultaneously developing its own AI shopping tools while blocking external AI agents through litigation. Control the agent layer so the advertising layer survives. But that's a holding action against a structural shift, and the timeline is deceptive.
Ad revenue is still growing. The three largest digital ad platforms together capture over 62% of global spending, and AI is actually improving their targeting in the short term. Quarterly earnings still look healthy. Meanwhile, the infrastructure being laid underneath points somewhere else entirely: payment protocols that don't need a browser, commerce standards that don't need a storefront, purchasing flows that never encounter an ad. The platforms that figure out what value they provide when nobody's watching are the ones that will still matter. The ones still optimizing for eyeballs may find themselves hosting a storefront where the customers have already sent someone else to pick up the order.