The website is no longer the primary interface

Ecommerce was built on a simple assumption.

Users discover products on your site.

That assumption is ending.

Discovery is moving into three external systems. Answer engines — ChatGPT, Gemini, Perplexity — that resolve product questions before a store visit. Aggregators — Google Shopping, Amazon, marketplaces — that compare and filter without requiring a click. Communities — Reddit, TikTok, niche forums — where recommendations form and trust is established.

By the time a user reaches a store, they are no longer searching. They are verifying.

This changes the role of the website completely.

The four jobs that became one

Historically the site performed four jobs. Attract — bring people in through search, social, ads. Educate — explain the product, the brand, the value proposition. Compare — let users evaluate options across the catalogue. Convert — close the transaction.

Now it performs one. Confirm.

The other three jobs have migrated. Attraction happens through entity presence in AI systems and marketplace feeds. Education happens through video, community, and AI-synthesised answers. Comparison happens on aggregator platforms and in AI responses.

What remains for the site is the confirmation step. The user arrives with intent already formed. They need to verify price, availability, shipping, trust signals. Then they either buy or leave.

Why metrics deteriorate while commercial health is stable

Traffic quality appears worse because the visit is shorter. In reality the visit is later in the decision timeline.

Pages per session decline because the user doesn't need to browse. Session duration drops because they already know what they want. Add-to-cart rate falls because the exploratory sessions — the ones that never converted anyway — have disappeared from the denominator.

But average order value holds or increases. Return rates stay stable or improve. Revenue per session often increases even while conversion rate drops.

The dashboard shows decline. The business is healthy. The measurement system is reporting a change in visit composition, not a change in commercial performance.

The optimisation mismatch

Most optimisation strategies still treat the site as a discovery environment.

More content pages. More category navigation. More exploration paths. More internal linking to keep users engaged. More on-page persuasion to "move users down the funnel."

But a confirmation interface needs the opposite.

Fewer choices. Stronger signals. Faster certainty. Definitive answers to the three questions every late-stage buyer has: is this the right product, is this the right price, and will this arrive when I need it?

The stores still optimising for discovery are adding complexity that slows confirmation. More elements on the page means more cognitive load. More navigation options means more decision branches. More content means more opportunities for contradictory information.

The commercial constraint has shifted

The constraint is no longer traffic acquisition. Most stores have sufficient traffic. The constraint is confidence resolution speed — how quickly the site converts a visitor who has already decided to buy into a completed transaction.

This is an architecture problem, not a design problem.

Confidence resolution requires consistent product data across every surface. It requires definitive inventory and shipping information at the product level, not at checkout. It requires removal of ambiguity, not addition of persuasion.

Stores that understand this remove friction. Stores that don't add persuasion. One scales. One plateaus.

What this means for the stack

If the website's primary role is confirmation rather than discovery, the infrastructure priorities change.

Product data quality becomes more important than content production. Feed architecture becomes more important than on-page SEO. Inventory accuracy becomes more important than merchandising creativity. Checkout reliability becomes more important than homepage optimisation.

The growth model shifts from "attract and convert" to "be present everywhere, confirm efficiently." The site becomes the transaction layer in a distributed discovery system, not the discovery system itself.

This is why conversion optimisation is increasingly an architecture problem, not a design problem. The constraint isn't the page. It's the system the page operates within.