1,300 semantic cocoons later: what I really learned about e-commerce SEO

Your competitors have an SEO asset that you are not exploiting. And it has a name — their brand

1,300 semantic cocoons deployed on e-commerce sites reveal four counterintuitive lessons about SEO: architecture takes precedence, total traffic is a bad indicator, authority determines

In 2023, I produced my three hundredth semantic cocoon. In 2025, the thousandth. Today, I count more than 1,300, deployed on e-commerce sites of all sizes, all sectors, all CMS.

At this volume, something changes. Intuitions turn into data. Particular cases reveal their regularities. And certain convictions, seemingly solid, crumble upon contact with reality.

Here are the lessons I couldn’t have articulated until I did this work at scale.

Architecture always takes precedence over content volume

The usual reflex of the e-retailer who wants to improve his SEO is to produce more. More articles, more pages, more words. This is rarely the problem.

In almost all of the cases I have handled, the blockage was not the lack of content, but the lack of structure between existing content. Isolated pages, without semantic link, without readable hierarchy. Google didn’t know what the site was the expert on.

Likama is the clearest illustration of this. 4,000 organic sessions per month, a catalog of 800 references, a technically correct site. But zero semantic architecture. Each product sheet lived in its own corner. We have built 23 semantic cocoons around the main purchasing intentions. Result: 37,000 organic sessions 14 months later, an increase of 820%, without additional advertising budget, without technical overhaul.

The content wasn’t the problem. The structure was.

Total traffic is a bad indicator

Here’s a truth that few consultants have any interest in telling: increasing the total traffic of a site does not necessarily generate more turnover.

GreenClosure sold fences. Their traffic came mostly from informational queries — people looking to understand how to install a fence, not how to buy one. We isolated the 47 queries with explicit commercial intent and restructured the architecture around them. Overall traffic has not exploded. Qualified traffic increased by 340% in six months. And organic turnover with it.

The relevant question is not “how many visitors?” but “which ones, with what intention?”

Speed ​​of results depends on starting authority, not content quality

One of the most common questions is: “How soon will I see results?”

The honest answer is: it depends on what Google already thinks of your site.

A site with low domain authority, few inbound links, short history — this one will wait 9 to 14 months before seeing stable positions. A site with 10 years of history and a solid link profile can observe significant movements in 3 to 4 months on the same structures.

Content is a necessary condition. It is not enough alone. The pre-existing authority is the multiplier.

Internal networking is the most underused lever

If I had to choose a single intervention on a site that is stagnating, it would be internal networking. Not the title tags, not the meta descriptions — the mesh.

A semantic cocoon only draws its power if the links between pages are thought in both directions: from the pillar page to the satellite pages, and from the satellite pages to the pillar. This circuit gives Google a clear signal: all these pages form a coherent whole on a specific subject.

Cabania, an e-retailer specializing in the rental of unusual cabins, had decent SEO. The overhaul of the internal network on existing pages, combined with three new cocoons in the most searched regions, produced an increase of 107% in six months.

No lines of code changed. No redesign. Links, structures, coherence.

What LLMs have changed

Since 2024, an additional observation is necessary: ​​semantic cocoons are naturally well positioned for the era of generative engines.

ChatGPT, Perplexity, Google SGE — these tools look for a source that covers a topic comprehensively. This is precisely the definition of a well-constructed cocoon. A site which answers sixty structured questions about unusual cabins in France will be more easily cited as a reference than a site which has one product sheet per accommodation.

Optimization for LLMs is not a revolution. This is the logical culmination of what quality SEO has always produced: real topical authority.

Leave a Reply

Your email address will not be published. Required fields are marked *