This article is extracted from Olivier Andrieu's conference at SEO Camp 2017. Olivier who wanted to thank above all the following SEO popes, who participated in the development of his slides:

The slides will be available soon and I will of course add the link here, but as we can't wait to share, here is the main part of this conference.

Presentation of the Semantic Cocoon

The principle: current websites are built on a tree structure based on the offer.

Example: I have a website on which I sell glasses. I have organized it with the following categories :

  • standard glasses
    • man
    • woman
    • child
  • shades
  • contact lenses

And the direct consequence, if I have a big site, is the mega (indigestible) menus, cf this screenshot of cdiscount :


but hey, it's not very user-based, is it? It's me, me and me again... Yet Google is trying to respond to the demand of Internet users, right? And we SEOs should too.

Well, the semantic silo is based on the request of the Internet user, so we reverse it: it's

a siloing technique based on the Internet user's request, with a linking strategy via an internal meshing of the pages between them. This affects both the content AND the internal mesh.

And what does it look like? I allow myself a screenshot from


We see a beautiful father, children, grandchildren ... it's good, isn't it? :)

4 steps to set it up

1st step: We classify the requests by the needs of the Internet user

What does my visitor want? I will classify my queries according to the visitor's requests:

  • Buy Eyeglasses
    • first purchase
    • frame renewal
    • desire to change
    • broken glasses
  • Eye protection
    • Sun / UV
    • From Insects.
    • From Projectiles
  • Eyeglass Reimbursement
    • Eyeglass prescription
    • ENT Appointment
      • ENT list by department
    • Mutual Reimbursement

Did you say demand and not supply? Gloups, we are far from my Cdiscount menu above... :)

Step 2: Definition of the silo (cocoon)

We take the 1st step, and we "map" it with the organization of the cocoon with target pages, intermediate pages (or "mixed"), and "final" pages (or "complementary").

  • Target Page: Eyewear
    • "Intermediate" or "blended" pages, which aim to:
      • push the upper target page,
      • suck in the lower pages
      • There can be several levels of intermediate pages
      • And there are finally:
        • "final" or "complementary" pages, whose objective is not necessarily to position themselves, but rather to push the content of the upper pages.

With this, we have created a "mindmap" per silo, which we will now have to implement!


Step 3: Create links

  • the target page links to the intermediate pages
  • that link to the intermediate pages (2nd category)
  • that link to the final pages (so, a whole bunch of "mother-daughter" links)
  • and...
  • we also make daughter-mother links, so we go up the tree at all levels!
  • and also "sister-sister" pages, whatever the level of the tree structure

Step 4: On-page optimization and internal linking

  • Link to top page
  • h1: Title
    • h2: subtitle
      • h3: paragraph with a link to a daughter page
      • h3: other paragraph with a link to another child page or intermediate page
      • To know more... :
        • Link to intermediate-sister page1
        • Link to intermediate-sister page2
        • Link to intermediate-sister page3

and the final page:

  • link to top page
  • h1
    • h2
      • h3 without any particular link :)
      • h3 without any particular link :)
      • For more informations :
        • links to sister pages


Well, this is of course only an introduction, and apparently Laurent Bourrelly is working (again!) on an improvement of the cocoon, so we remind you that this is really "only" an introduction!

More details here :


Principle: a query corresponds to a search intention, which corresponds to a metaword (footprint / signal).

1 metaword is represented as a set of lexies (expressions) that will be used in the text of the page. The lexies can be more or less attracted to each other and that's the magic :).

More details here:

Example: the metaword "evolutionary algorithm" will have the following lexies:

  • algorithm
  • optimization
  • selection
  • evolution
  • function
  • population

which each have a degree of attraction towards the metaword. This metaword is created on the basis of the analysis of the web pages corresponding to the request (not stupid, since these are the pages that rank ;)). The choice of these lexies is therefore fundamental according to the attractiveness score:

  • - above 9000: mandatory because omnipresent
 - between 2000 & 8000 : determining, and therefore very important
 - below 2000: useful!
  • >2000 & <8000 : déterminants, et donc très importants
  • <2000 : utiles!

These lexies are fully compatible with the semantic cocoon: it can work between 2 linked pages to help Google understand the content of the destination page.

And more...

We had a quick presentation of the following tools:


Now you know what a cocoon is and how to create one.

   Article written by Louis Chevant

Further reading

The complete guide to Internal Meshing

The step-by-step method to build your semantic cocoons, your mesh and the optimal tree structure of your website.