Assistants Are Answering Category Questions With Shortlists — and There Is No Page Two

Assistants have started answering category-level shopping questions with a short list rather than a wall of links. Ask which mid-range espresso grinder to buy, and the reply increasingly names two or three models, states why each was included, and stops. The notable change this quarter is what sits behind that selection: the answer is assembled from a smaller number of pages than a traditional search results page, and those pages tend to be the ones that state their attributes in extractable text.

That has a practical consequence for merchants. When an assistant narrows a category to three candidates, the cost of being fourth is not a lower ranking — it is absence. There is no page two. A store that ranks comfortably in conventional search for a category term can still never appear in an assistant's shortlist if its key attributes are published as images, or stated differently on the product page and the feed.

We are also seeing the questions get more specific rather than less. Buyers ask about warranty length, filter compatibility, restock timing, and whether a size runs large. Each of those is a factual claim that either exists in your text or does not. Assistants do not infer a measurement from a photograph, and they do not average conflicting values across your own pages — they pick a source that is unambiguous and quote it.

## What this means in practice

The merchants appearing in these shortlists share a pattern rather than a tactic. Their product pages answer a defined set of questions in plain text, their structured data agrees with what is visible on the page, and their feed matches both. None of that is exotic. It is the same consistency discipline that governs good catalog hygiene, applied to a retrieval process that reads rather than ranks.

The work is also smaller than it sounds. Most audits find a short list of blockers: a handful of attributes living only in images, a few conflicting values between page and feed, and missing markup on question-and-answer content. Fixing those is a matter of hours, not a replatforming project.

## The Visora angle

We built the free scanner at /audit to surface exactly this — which attributes are missing, malformed, or conflicting, ranked by how often each class blocks citation on the questions buyers actually ask. It is diagnostic, not a guarantee, and no tool can promise placement in an assistant's shortlist. What it can do is tell you, in one pass, whether your pages are readable and consistent enough to be a candidate at all.

If you are not sure where you stand, scan your three highest-margin products first. The findings are usually representative of the whole catalog, and they are usually fixable this week.