Assistants do not rank pages, they assemble answers. That changes what a catalogue has to be good at: not winning a position, but being the most quotable source of a specific fact. Here is what that looks like in practice, and what it does not.
The Dazzle Commerce Team
eCommerce Management & Marketing
Updated
5 September 2026
Reading time
7 min read
Assistants answer in sentences. The question is whose sentence gets used.
The honest starting position: nobody outside these companies knows exactly how an assistant picks its sources, the systems change frequently, and anyone selling certainty about it is selling something. What is observable is the shape of the problem, and the shape has not changed much.
An assistant is trying to answer a question in a paragraph. To do that it needs facts it can state without hedging, attributed to a source it can defend. A catalogue that supplies clean, specific, verifiable facts is easier to quote than one that supplies adjectives.
KEY TAKEAWAY
Search asked: is this page relevant? An assistant asks: can I safely say this sentence, and who do I credit for it? Optimise for the second question.
Across the surfaces we watch, four things consistently separate a cited source from an ignored one.
1
The fact is stated plainly and in one place. A specification split across a tab, an image and a PDF is three facts an assistant has to reconstruct. One table is one fact it can lift.
2
The claim is specific enough to be checkable. “Water resistant to 50 metres” can be quoted. “Built to handle whatever you throw at it” cannot.
3
The source is corroborated elsewhere. The same specification appearing on your site, your marketplace listings and your retail partners reads as a fact. Appearing once reads as a claim.
4
The page is reachable and parseable. Content that only exists after a click, a script or an interaction may as well not exist for a system reading the page.
Markup does not persuade anything to cite you. It removes the ambiguity that stops you being cited. Get it complete and consistent, then stop thinking about it.
Signal
Why it matters
Where it belongs
Product identifiers
Lets a system match your item to the same item elsewhere
Product markup and every feed
Price and availability
Assistants avoid quoting facts that go stale
Product markup, kept in sync
Specification attributes
The raw material of a comparison answer
A real table, not prose or an image
Review data
Supports the judgement part of an answer
Review markup plus visible reviews
Breadcrumb and category
Establishes what kind of thing this is
Breadcrumb markup on every template
Consistency across your site, your feeds and your marketplace listings matters more than any single implementation detail.
This is the part most teams skip. If you want to be the source for “which of these is best for narrow feet”, somewhere on your site there has to be a clear sentence that answers it. Four habits get you there:
The practical test: could someone answer a shopper’s question by quoting one sentence from your page, without needing to interpret it?
Star averages are almost useless to an assistant. What it can use is a review that names a situation: a runner with a wide foot, a kitchen with limited counter depth, a first-time buyer. Review volume without specificity is noise; a smaller set of detailed reviews is a richer source.
per fact
Specifications reconstructed from three places rarely get quoted
across every channel
Site, feeds and marketplace listings should agree exactly
answers the question
Content that builds to its point loses the point
Working principles from our own catalogue and content work rather than published research.
The practical move is to change what you ask for in post-purchase review requests. Asking “how did you use it and who would you recommend it to?” produces materially more useful text than asking for a rating.
On Amazon the assistant is reading your listing, your A+ content, your attributes and your reviews — a closed set you control almost entirely. That makes it the most tractable of these surfaces.
1
Fill every applicable attribute in the listing back end, including the ones that do not display. They are still structured facts about the product.
2
Write bullets that answer use-case questions rather than restating features. “Fits standard 60cm cabinets” beats “compact design”.
3
Treat the customer questions section as content. It is a direct record of what shoppers ask, and answering it well feeds the same system.
EXPERT VIEW
“The work that makes a catalogue quotable is the same work that makes it clear. We have not yet found an assistant optimisation that was not also a plain improvement to the page.”
Dazzle Commerce · Organic search team
Catalogue consistency across site, feeds and marketplaces is a substantial piece of work. It is one of the things our SEO and marketplace teams do jointly.
There is no rank tracker for this yet, and the reporting is genuinely immature. Three things are worth watching anyway:
A SENSIBLE POSITION FOR NOW
Do not rebuild a content strategy around this yet. Do make your specifications complete, consistent and quotable — that work pays in ordinary search regardless of what assistants do next.
Three services that make a catalogue clear enough to be quoted.
Structured data, category architecture and content built around the questions people actually ask.
Product copy and comparison content written to answer a question in its first sentence.
Listing attributes, A+ content and review programmes on the surface where Rufus is reading.
WRITTEN BY
eCommerce Management & Marketing
Dazzle Commerce runs organic search and marketplace content programmes for eCommerce brands. This piece reflects how we are currently adapting catalogue and content work for assistant-led discovery, and where we think the honest limits of that work sit.
More from the organic search and catalogue side of the practice.
The page layer that carries commercial demand in ordinary search as well as assistant answers.
The same attribute completeness problem, measured in wasted ad spend.
Complete specifications help shoppers and machines for exactly the same reason.
We will run your top categories through the same checks we use on client catalogues and show you where the answers come from instead.
