That 10% Revenue Stat? The Vendor Selling It Told You So

If you sell things online and an AI agent is the customer, congratulations — you probably don't know how to be found. That's the actual story buried under the Fortune headline about agents driving 10 percent of revenue for some brands. The 10 percent figure is self-reported by clients of a vendor that sells agentic optimization services, which is exactly the kind of evidence you'd reject in any other context. The more durable, verifiable signal is this: as AI agents become a real traffic source, the infrastructure for being discovered by them is largely nonexistent for most brands.

McKinsey projects that agentic commerce will drive up to $1 trillion in U.S. retail revenue by 2030, according to a March 2026 Fortune column by Aviv Shamny, CEO of Limy. That's a projection, not a measurement, but it's the right order of magnitude for something that represents a meaningful share of a $5 trillion sector. And the early read on where agents are actually sending people suggests the winners in this new distribution layer won't be the brands that dominated Google search.

Target's traffic from ChatGPT is growing 40 percent month over month. Walmart has seen ChatGPT referral traffic reach as high as 20 percent of its referral visits. These aren't edge cases — they're the leading edge of a pattern. An Adobe study found that 14 percent of U.S. consumers already rely on ChatGPT over Google for product discovery. That's not a rounding error.

The gap becomes starker when you look at where agents actually find things. A Semrush study of 15,000 ChatGPT prompts found that 90 percent of the sources the model cited were not on Google's first 20 pages. An Ahrefs study of 15,000 prompts found only 12 percent overlap between URLs cited by AI tools and the top 10 results in Google search. The implication is uncomfortable for any brand that built its digital presence on SEO: the agentic web has a different geography, and it wasn't mapped by the same forces that determined PageRank.

This is where the infrastructure story gets interesting. Agentic discovery operates on a fundamentally different lookup model than search. A traditional search engine crawls publicly accessible pages and ranks them by links and relevance signals. An agentic system — depending on how it's designed — may query product databases, read structured data feeds, pull from knowledge bases, or synthesize responses from sources it's been trained on. The optimization surface isn't a webpage with good metadata. It's an ecosystem of APIs, structured product feeds, and knowledge graph entries that most brands haven't thought about since the early 2010s.

Some companies are already moving to fill this gap. Limy — whose CEO wrote the Fortune column that prompted this story — is one of a handful of startups building what they're calling "agent experience optimization" (AEO) services. Limy emerged from stealth in January 2026 with $10 million in funding. The pitch is roughly analogous to SEO: audit your brand's presence across the systems agents actually query, restructure product data to match the retrieval patterns agents use, and measure visibility the same way you'd measure search rankings. One robotics customer, unnamed, achieved a 94 percent increase in agentic visibility in four months through AEO restructuring. Whether that translates to revenue is a different question — the 10 percent revenue attribution claim comes from the same clients Limy is paid to optimize, which is not a clean evidence chain.

The irony here isn't subtle. SEO was the infrastructure layer that let the commercial web scale — it gave brands a way to be found before the transaction happened. A similar layer is emerging for agents, and it's being built by some of the same playbook writers, with some of the same marketing language, and some of the same conflicts of interest that made early SEO a Wild West of snake oil and real results living side by side. The difference is that the agentic lookup model is more varied and less standardized than Google ever was, which means the optimization play is harder to get right and easier to sell as magic.

There are also the obvious questions that a self-interested vendor won't answer in a Fortune op-ed. How much of agentic product discovery is going through proprietary agent systems — retail chains' own recommendation agents, vertical SaaS platforms with embedded AI — versus the consumer-facing chatbots everyone talks about? What does the attribution model even look like when an agent synthesizes information from ten sources and the user never clicks through to a brand's site? And at what point does "we increased your agentic visibility by 94 percent" become a metric that's structurally impossible to verify independently?

These aren't reasons to ignore the shift. McKinsey's trillion-dollar projection may be aggressive or conservative — it's in the right range to matter. Target and Walmart's ChatGPT referral numbers, while imperfect, point to real volume moving through real channels. The question for brands isn't whether agentic discovery is coming. It's whether the infrastructure to be found in it is something they need to build, buy, or wait for their platform vendors to handle.

Our read: the brands that figure this out first will have a meaningful advantage in the next phase of commercial AI. The brands that treat it as a future problem are betting that their existing SEO moat translates — and the Semrush and Ahrefs data suggest it doesn't, not cleanly. The plumbing is different. The lookup graph has a different shape. And the companies building the pipes are, mostly, the same companies selling the shovels.