The part of a purchase that used to send people to a dozen pages is increasingly answered in a summary at the top of the results. That is the change worth planning around, and it is not really about a feature. It is about which stage of the buying process still produces a visit.
Research questions produce fewer clicks now, because a competent summary answers them. Decision questions still produce clicks, because at some point a person has to see the actual product, the actual price and the actual delivery date, and no summary can supply those.
So the value of a page has shifted according to where it sits in that process. A general explainer that competes on being a tidy summary is doing work that is now done above it. A page that carries specifics only you have, your stock, your dimensions, your compatibility, your policies, is doing work a summary can only cite.
Our reading is straightforward. Write less to be read, and more to be quoted.
Being cited is the new first position
A summary is assembled from sources, and the sources get named. That citation is a small, high-intent doorway, and it behaves differently from a normal ranking. It arrives with the shopper already partly convinced, because your page was used as the evidence for an answer they just accepted.
What gets cited, in our experience of watching this, is text that answers a question in a self-contained way. A paragraph that states a fact, plainly, with the qualifier attached, is quotable. A paragraph that builds atmosphere for four sentences before arriving at the point is not.
That has a direct consequence for how pages should be written. Lead every section with the answer, then explain. Keep each section able to stand alone, because it may be lifted alone. Put numbers, dimensions, timelines and conditions in the text rather than only in an image or a table image.
It also raises the value of specificity. A claim that is true of every store in your category is not worth citing. A statement about how this exact product behaves, from a store that clearly sells it, is.
Write less to be read, and more to be quoted.
What to write now that summaries answer the basics
Stop writing the article that explains the category in general terms. That page competed on organisation, and organisation is what these systems are best at.
Write instead what a summary cannot generate. Comparisons between your own options, with the trade-off stated. Fit and compatibility detail for the products you actually stock. What goes wrong with this type of product and what to do about it. Care, maintenance, and how long things last in real use. The questions your support inbox answers all day, with the real answer rather than the safe one.
All of that has a second benefit. It is the same material that closes the sale on the product page, so the work pays twice. The store gets a page worth citing and a page worth converting on, which is a better use of a writing budget than a blog post aimed at a keyword nobody clicks any more.
And keep it attached to commerce. A page that answers the research question and then shows the products that satisfy it is doing the whole job. A detached article that sends the reader off to search again is doing half.
Technical basics still decide whether you are eligible
None of the above matters if the page cannot be read cleanly. The unglamorous groundwork is the same groundwork that has always mattered for search, and it is now the price of entry.
Product pages need accurate structured data: price, availability, brand, identifiers, review data if you have it and it is genuine. This is how a machine confirms the specifics it wants to quote, and a mismatch between your structured data and your visible page is a reason to distrust the page.
Content needs to exist in the HTML rather than appearing only after a script runs. Text hidden behind tabs or accordions is usually fine, text that only exists after an interaction is riskier.
Pages need to be reachable and reasonably fast, and the important content should not be an image of text.
And your facts need to agree with each other. If the page says one delivery time, the policy page says another, and the feed says a third, you have taught every reader, human or otherwise, that your information cannot be relied on.
You cannot measure this the old way
The reporting problem is real and there is no clean fix, so it is better to change what you look at than to wait for a perfect metric.
Impressions can rise while clicks fall, and that is not automatically a failure. It can mean your page is being used in an answer by people who did not need to visit. It can also mean you lost the click to someone else. The figure alone will not tell you which.
What we watch instead is the quality of what arrives. Sessions from search that reach a product page, add-to-cart rate on those sessions, and revenue per session from organic search. If fewer people arrive but more of them buy, the research traffic you lost was traffic that was never going to buy today.
We also watch brand searches. When a store starts being named in answers, the effect often shows up first as people searching for the brand directly rather than as a rise in clicks on the original page.
And we test the questions by hand. Ask the assistants the questions your customers ask, once a month, and write down whether you are mentioned and what is said. It is slow and manual and it is the most reliable signal available.
What we would do this quarter
Four moves, in order, and none of them require abandoning what already works.
Pick the ten questions your buyers actually ask before purchase, taken from your support inbox and your on-site search, and answer each one properly on a page that also sells the relevant products.
Rewrite the openings of your best-performing existing pages so the answer comes first and each section stands alone. This is editing, not writing, and it is the fastest of the four.
Audit the structured data on your product pages and fix any disagreement between what the page shows and what the markup says.
Build the small monthly habit of asking assistants your own buying questions and recording the results, so that in six months you have a trend rather than an opinion.
What we would not do is stop investing in search because of the change. The stores that are cited are, overwhelmingly, the stores that had good pages already.