A shopper hunting for a replacement water filter used to type a few words into a search box, scan a page of blue links, click the one that looked closest, and land on a category page that did the real selling. The category page sorted the options, surfaced the specs, and nudged the reader toward a product.
Now, the same shopper reads a synthesized answer at the top of the results, sometimes clicks nothing, and if they do click, they arrive half-decided. Pew Research found that users clicked a traditional result on 8% of visits when an AI summary was present, versus 15% without.
The category page is no longer the first thing read. Often it’s the second, if it gets read at all. That changes what you should put on it, and what you should stop spending on.
Contents
Decide If Category Pages Are for Humans or Answer Engines
The honest answer is both, but not in equal measure, and the proportions have moved. A category page used to be judged by how well it helped a browsing visitor narrow a set. That visitor still exists, but a second reader has shown up: the model that scrapes the page to assemble an answer before anyone clicks.
You have two real choices. Keep writing merchandised prose that reads well to a human and hope the model picks up what it needs. Or restructure the page so the first screen answers the obvious question in plain language, with the merchandising underneath.
The second approach costs more up front and asks writers to think like editors, but it’s where most teams investing in search and answer-engine work are heading. If your category pages still open with a hero image and three lines of brand voice, you’re writing for a reader who may never see them.
Decide What the First Screen Has to Earn
The first screen is now doing two jobs at once. It has to satisfy a model looking for an extractable answer, and it has to reward a human who arrives already informed and impatient. Write the top of the page the way a careful editor would answer out loud if someone asked what this category is and how to choose inside it.
That means leading with the distinctions that actually affect the purchase: who the products are for, what separates one option from another, and which specifications matter most. Save the brand language and merchandising flourishes for farther down the page. The first screen has to earn enough understanding that a model can summarize it and enough confidence that a shopper keeps going.
Decide Which Pages to Rewrite and Which to Retire
Most catalogs carry more second-tier pages than anyone wants to admit. Subcategory pages built for a taxonomy that made sense in 2019. Tag pages the CMS created on its own. Filter combinations that spawned their own URLs.
The question to ask about each one is simple: if a model were summarizing this topic, would this page add anything the main category page doesn’t already cover? If the answer is no, consolidate it. Google’s own ecommerce URL guidance is worth re-reading before you start merging, because canonicalization and pagination choices determine whether the surviving page inherits the authority of the ones you fold into it.
A thinner site where each page says something distinct tends to get cited more than a sprawling one that spreads the same information across twelve URLs.
Decide How to Measure a Page Nobody Clicked
The old scorecard was traffic and conversion rate. Both still matter. Neither tells you whether your category page is doing its new job, which is being read inside an answer the user never clicked through.
Add two things to the dashboard. First, track citations and mentions of your pages inside answer-engine results, not only the clicks that come back. Second, watch what happens to downstream conversion when direct and branded traffic rises while category-page sessions fall.
That pattern usually means the page is being read upstream and the buyer is arriving ready. Count sessions alone and you’ll kill the page that’s actually working.
Decide Who Owns the Rewrite
Category pages have historically belonged to merchandising or e-commerce, with SEO consulted. That split no longer fits the work. Rewriting these pages for extractable answers is an editorial job as much as a merchandising one, and the person doing it needs to care about how a sentence will read when a model lifts it out of context.
Pick an owner by name. Give them the authority to shorten brand copy, add plain definitions, and consolidate pages that duplicate each other. Then give them a review cycle, because answer engines keep changing what they surface, and the page that gets cited this quarter may not be the one that gets cited next.
The teams that move first will have cleaner pages, fewer of them, and a measurement setup that reflects what the second-tier page is now being asked to do.
