Keyword search just died, and five CEOs are the proof.

A CEO doesn’t usually confirm a competitor’s thesis by accident. This month, five of them did, independently, across five different companies that don’t coordinate talking points: a payments network, a search giant, a commerce analytics firm, a product-data infrastructure company, and the platform running more storefronts than anyone. None of them were talking to each other. All of them said the same thing, and they said it in earnings calls, research reports, and press briefings that had nothing to do with one another.

Let’s take a closer look at these, because it’s easy to read any one of these claims as a vendor talking its own book. A payments company saying agentic commerce is big is expected. A search company saying AI matching matters is expected. What’s harder to wave off is five of them, across the full stack from payments to search to catalog infrastructure, landing on the identical read in the same stretch of weeks, with the receipts to back it.

The numbers stopped being soft

Start with what changed on the ground. Nearly 80% of commerce professionals now report a measurable traffic increase from LLM-powered search. Ninety percent expect AI search to be essential to product discovery within the next 12 months, not eventually, within a year. Seventy-one percent already call AI core to how they operate today, not a pilot program. Consumer use of AI shopping assistants is up 200% since last year alone.

The flip side of that growth is where it’s coming from. Brand-owned product discovery, a merchant’s own site doing the finding, fell 7%. Traditional keyword search dropped 15%. That’s not two numbers moving independently. That’s demand physically relocating from the channels merchants control to the channels they don’t, in the same twelve months the adoption numbers above were climbing.

Multi-dimensional matching favors the long tail, not the giants

Visa’s CEO framed the shift in market-expansion terms this week: agentic commerce, in his words, expands the addressable market, and the company is still in the early stages of capturing it. That’s a payments executive talking about the same mechanism that matters to a small merchant, just from the other end of the transaction.

AI agents don’t rank by keyword match, they match on multiple dimensions at once, color, fit, price, availability, occasion, all evaluated together instead of one query term at a time. That kind of matching structurally favors the long tail, the specific product for the specific buyer, over the generic bestseller that used to win a keyword auction. A search paradigm built on exact-phrase matching rewards whoever bid highest on the phrase. A search paradigm built on genuine relevance has no particular reason to prefer the giant over the specific, well-matched answer, wherever it lives in the catalog.

Two-sided disruption, not one

Salesforce’s read on this, from its own State of Commerce research, is the sharpest one-line summary of why this is harder than a normal channel shift:

AI remakes commerce from both directions at once, because customers are now discovering products through channels the brand does not control.

That is a different problem than “our SEO ranking dropped” or “our paid CPCs went up.” Those are competitive problems inside a channel merchants already understood. This is the channel itself being rebuilt by someone else’s model, on someone else’s interface, using product data the merchant supplied but does not get to curate the presentation of.

Google is leaning into exactly that position rather than away from it. Its Universal Cart AI shopping assistant and the Gemini models underneath it are explicitly built to sit between the shopper’s intent and the merchant’s catalog, and Google’s own leadership describes the value proposition in matching terms, not search terms:

The model super-charges the ability to understand what someone is actually looking for and match it correctly.

That is a search company quietly admitting that the keyword box it built an advertising empire on is not the primary interface anymore. The matching layer is.

A trillion listings a month is a new kind of infrastructure claim

This is where the infrastructure argument stops being a talking point and starts being a number. A product-data platform processing north of a trillion product listings a month isn’t a nice-to-have feature vendor, it’s plumbing at the scale of the commerce internet itself, and its CEO is now describing structured product data in exactly those terms: foundational infrastructure for modern commerce operations, not a back-office data hygiene task anyone gets to defer.

Shopify’s own president made the same point from the merchant side, in plainer language: in this world, relevancy reigns, and specific products matched to specific buyers are the ones that perform. Half of AI-referred sessions now land directly on a product page instead of a search results page. That’s not a traffic story. That’s a data-readiness story wearing a traffic story’s clothes. A product page can only be the landing spot if the underlying catalog is structured well enough for an agent to resolve intent straight to it, no browsing, no filtering, no guessing required.

Put the infrastructure claim and the merchant claim next to each other and they’re describing one system from two ends: the platform that indexes at trillion-listing scale, and the storefront where relevancy either shows up or doesn’t. Everything in between, the actual attribute quality, the taxonomy, the descriptions an agent can actually parse, is the part most merchants still haven’t touched.

Early innings, and the caveat that matters

None of this should be read as settled. The standards are still competing, the winners aren’t decided, and buyer beware still applies to any vendor promising a finished category this early. That caution is fair and worth keeping.

But “early innings” and “wait and see” are not the same posture, and the gap between them is exactly where most merchants are currently standing. The adoption curve above isn’t a forecast anymore, it already happened. The traffic already moved. The question a merchant on Shopify, BigCommerce, or WooCommerce actually faces isn’t whether this is real. Five unrelated executives just spent a week telling anyone paying attention that it is. The question is whether their own catalog is one an agent can actually read, match, and land a shopper on, or whether they’re the 7% and the 15% still waiting to notice the floor moved.

To all merchants and the agencies that support them – It’s time to wake up.

Main street commerce, the tens of millions of stores that aren’t running trillion-listing infrastructure budgets, doesn’t get to wait for the standards fight to resolve before fixing its own data. The merchants processing that trillion-listing scale, the Shopifys and Googles and Visas of the world, already have the engineering headcount to instrument for this on their own timeline. The merchant running a single Shopify Plus or BigCommerce storefront does not, and nothing in this week’s numbers was built with that merchant in mind. That gap between what the platforms are proving works at their scale and what an individual merchant can actually afford to implement at theirs is the one still mostly unaddressed, and it’s the one worth building for now, not after the caveats clear.