SEO for product sites: structure beats volume
Search & Tooling · 2026-06-28 · 6 min read · 1380 words
By ESPYCRUX, Studio
Search now often resolves without a click, and generated answers cite two or three sources rather than listing ten. Being the fourth-best page used to earn traffic; it now earns nothing. Here is what actually still works for a small product site, and what to measure when the click never happens.
Most SEO advice was written for a search results page that behaved like a list. Ten links, ranked, and your job was to be higher up the list than you currently were. Volume worked, because more pages meant more entries in more lists.
That page has changed shape twice. First the snippet took the click for simple questions. Then generated answers began compressing several sources into one paragraph with a handful of attributions. Both changes reward the same thing and punish the same thing, which at least makes the response simple to describe.
What is punished: being adequate, at volume, on topics many people cover. What is rewarded: being the source worth quoting.
Structure is the part most sites skip
Before anything about content, a product site needs a shape that explains itself.
A small number of pillar pages — the things you actually make — each surrounded by specific supporting articles that link back with descriptive anchors. Not fifty scattered posts. Not a blog that exists parallel to the product and never touches it.
Three rules that do most of the work:
Every article links to the product it supports. If a piece is about browser performance and you build browser tools, the connection should be a link, not an implication. This is the mechanism by which writing supports a product rather than merely accompanying it.
Every article links to two siblings. It distributes authority across the cluster and keeps a reader moving instead of leaving. Internal linking is the highest-leverage thing most small teams ignore entirely, and unlike almost everything else in this article it is fully under your control.
Anchors describe the destination. "Read more" tells a crawler nothing. "Our Core Web Vitals checklist" tells it exactly what is at the other end.
The unglamorous items matter too: unique titles and descriptions per page, kept under the length where they get truncated. It is dull work and it moves click-through more reliably than another thousand words of body copy.
Structured data is no longer decoration
Organization, Article, BreadcrumbList and FAQPage markup give a search engine an unambiguous read of what a page is, rather than making it infer.
The reason to bother is that inference is exactly what gets skipped when an answer is being assembled from several sources at once. Explicit markup is cheap to maintain when it is generated from the same data the page renders — you are not writing JSON-LD by hand, you are emitting it from the object you already have.
One thing worth checking: markup that contradicts the visible page is worse than none. If your FAQPage markup lists questions a visitor cannot see, that is a problem, not an optimisation.
The zero-click problem, stated honestly
Search increasingly resolves on the results page. Someone gets the opening hours, the definition, the conversion, and never visits anyone. Your traffic chart can decline while your visibility grows, which is a genuinely difficult thing to explain to anyone judging you on sessions.
There are three responses, and the order matters.
Make sure the extracted answer carries your name. Accurate structured data, clear factual sentences and consistent business information all increase the chance that what gets shown attributes rather than anonymises. This is defensive and it is not nothing.
Move the valuable material one step deeper. Quick facts will be taken and displayed — that is now simply the deal. Judgement, comparison, methodology, tooling and original data cannot be compressed into a snippet without losing what makes them useful. Those are what a click is still worth making for.
Reduce your dependency on the channel entirely. A mailing list, a community, a product loop — anything nobody can restructure without your involvement. Every year makes this argument more persuasive, and every year more people wish they had started earlier.
Writing for answers that cite you
Generated answers cite a small number of sources. Being the fourth-best page earns nothing, while being the cited source earns the whole reference. That is a harsher distribution than ranking fourth on a list of ten, and it changes what "good enough" means.
Content that gets cited shares traits. It is specific. It commits to numbers and dates. It states its methodology. And critically, it is written in self-contained passages.
That last one is testable, and the test takes ten seconds:
Take any paragraph from your page. Paste it somewhere with no heading and no surrounding context. Does it still stand up?
If it needs the sentence before it to make sense, an assistant will struggle to quote it cleanly, and a reader arriving mid-answer will bounce. Writing that survives being lifted out of the page is a different discipline from writing that flows — and you want both, which mostly means putting the claim first and the elaboration after, in every section.
Machine readability is decisive again
This one is unglamorous and increasingly load-bearing.
Content that only exists after client-side JavaScript has run is harder to retrieve. So is content buried behind tabs, split across paginated fragments, or assembled from an API after load. Clean HTML, stable URLs, honest headings and accurate structured data are cheap, boring, and durable.
We prerender every page on this site, so the full text is present in the initial HTML rather than assembled by a script after load. That is not a growth tactic; it is the baseline that makes everything else legible to a crawler that may not run your JavaScript at all. If you have built a single-page application and your content only appears after hydration, this is the highest-value thing on this list and nothing else you do will compensate for it.
A related trap worth naming: do not gate content on scroll. If your reveal animations inject content on intersection rather than transforming content that is already present, a crawler that does not scroll sees an empty page. Transform, do not gate.
Measurement has to change
If you keep measuring sessions, you will conclude you are failing during a period when your visibility is growing. The metrics that still mean something:
| Signal | What it tells you | Why it survives zero-click |
|---|---|---|
| Branded search volume | People looking for you by name | The clearest evidence exposure worked, even with no click |
| Direct traffic | People arriving already knowing you | Often the shadow of a citation you cannot attribute |
| Referral mentions from assistants | Where you are being cited | Partial and inconsistent, but real |
| Impressions vs clicks | Visibility separated from traffic | Rising impressions with flat clicks is the zero-click signature |
| Sessions alone | Very little now | Understates presence whenever the answer never produced a click |
There is a practical tell worth watching for: visitors who arrive already knowing your numbers, your framework or your product name. Someone an assistant sent to you without a click you can attribute. Rising branded search and direct traffic while referral clicks stay flat is usually the shadow of citations you cannot see in your analytics.
The durable move
Everything above is tactics. The strategic point is shorter.
Own something nobody can synthesise. A model can rewrite an opinion into ten fresh sentences without needing to point at anyone. It cannot invent your data. Run the survey. Publish the benchmark. Share the number only you have. Describe the thing you actually built and what happened when you built it.
Aggregators get summarised away, because there is nothing in them that is not somewhere else. Primary sources get cited, because there is no alternative route to the information.
This is also, conveniently, the same advice as writing something worth reading — which is why it is one of the few SEO recommendations that will still be true after the next change to the results page.
The short version
- Small number of pillar pages, each with supporting articles linking back
- Every article links to the product it supports and to two siblings
- Descriptive anchor text, unique titles and descriptions
- Organization, Article, BreadcrumbList, FAQPage markup, generated from your data
- Full text in the initial HTML — prerender if you are client-rendered
- Paragraphs that stand alone when lifted out of the page
- Commit to specifics: numbers, dates, methodology
- Publish something only you have
- Measure branded search and direct traffic, not sessions alone
- Build one channel search cannot take away from you
Tags: seo, zero-click, structured data
ESPYCRUX — ESPYCRUX is a small product studio based in India, building focused web applications and writing about the engineering behind them. Articles are written by whoever did the work, and published under the studio name. Reach the studio at admin@espycrux.com.