Marie Haynes argues in Search Engine Journal that GEO hype is pulling attention away from the agentic web, where Universal Cart, UCP, WebMCP, and intelligent eyewear are creating the real near-term openings for businesses.
Marie Haynes argues in Search Engine Journal that generative engine optimization has become a distraction from a larger shift: the agentic web, where software agents transact and act on a user’s behalf. Her piece, published October 23, 2026, maps what she frames as opportunities most businesses are currently ignoring.
What the argument actually says
The substance of the piece is a redirect. Haynes treats GEO, the practice of optimizing content so generative engines surface and cite it, as a narrow concern relative to what she says is already arriving. She names four developments as the evidence: Universal Cart, UCP, WebMCP, and intelligent eyewear. Her claim is that these signal an agentic web in motion, and that businesses are spending their attention on GEO while leaving the openings those developments create unaddressed.
That is the whole of the reported position. The piece does not, per the available summary, quantify adoption, name specific vendors behind each technology, or offer a step-by-step playbook. It is an argument about where attention should go, backed by a list of technologies Haynes says are arriving now.
Why this lands differently for a site owner than for a marketer
GEO is a comfortable problem because it looks like SEO. You write, you structure, you measure citations, you iterate. The mental model carries over from a decade of search work, which is exactly why it absorbs so much budget and discussion. The agentic web does not offer that comfort. If an agent is completing a purchase, booking a slot, or assembling a shortlist on a user’s behalf, the thing being optimized is not a snippet. It is whether your product data, inventory, pricing, and action endpoints are legible to a machine that will never see your landing page.
That is a different discipline. It sits closer to product feeds, structured product data, and API design than to content strategy. A marketing team can own GEO. Nobody owns the agentic web inside most organizations, which is a decent explanation for why it gets deferred.
The four signals, read plainly
Universal Cart points at commerce: a cart that follows a user across surfaces rather than living inside one storefront. UCP and WebMCP point at protocol and tooling, the plumbing that lets agents discover and invoke capabilities on the web. Intelligent eyewear points at the interface layer, where an assistant sees what the wearer sees and acts inside that context.
Read together, they describe a web where the buyer is often software and the interface is often not a browser tab. For a developer, that means the interesting work is in machine-readable product and service data and in exposing actions an agent can call safely. For a marketer, it means the funnel may start inside an assistant with no impression, no click, and no session to attribute. For a site owner, it means the question shifts from “do we rank” to “can an agent complete this task against our systems without a human on our side.”
None of that makes GEO worthless. It makes it partial. Being cited by a generative engine is one way to be chosen. Being transactable by an agent is another, and it is the one with fewer people working on it right now.
What to watch next
The honest gap in the argument is proof. Haynes asserts the agentic web is arriving; the useful follow-up is evidence of volume. Watch for concrete adoption numbers around Universal Cart and the protocol work behind UCP and WebMCP, and for whether intelligent eyewear ships in quantities that matter to consumer-facing businesses rather than demo units.
Watch, too, for the standards question. Agent-facing commerce only works if agents can read the same signals across merchants, and that consolidation has not happened yet. Until it does, the practical move is unglamorous: audit whether your product and service data is clean and machine-readable, and whether the actions an agent would need to take are exposed at all. That work pays off under either scenario. Chasing GEO alone does not.
The sharper point in Haynes’s framing is about attention economics. A field with clear metrics and familiar workflows will always outcompete one that is still forming, even when the forming one matters more. That is not a failure of GEO practitioners. It is how teams allocate scarce time, and it is why the businesses that start on agent-readiness now will have a quieter advantage than the ones publishing another round of GEO experiments.
Source: searchenginejournal.com