Article · October 27, 2026 · 11 min read
Topic frame: how to design a semantic cocoon and its internal linking
Lily Ray observes that many sites that lost ground in 2026 grew by volume, with similar pages for every shade of a query. The alternative is a structure where each page has its own job and the links between pages follow a person’s path to a decision. French SEO calls this a semantic cocoon. We call our version a topic frame, and here’s how to build one in five steps.
Where the idea comes from and why “frame”
The semantic cocoon (cocon sémantique) was developed and popularized by French SEO Laurent Bourrelly. The structure is built around a target page: supporting pages reinforce it and link to it and to each other with contextual links, and the user’s search intent sets the logic. French writers sum up the difference from a silo briefly: the silo separates, the cocoon connects.
In Russian-language SEO, DrMax develops the approach further. He suggests building a cocoon on what people themselves write about the problem, not only on keyword clustering, and calls these fourth-generation cocoons. For building them he has a system of four roles: audience research, architecture, content logic and a coordinating layer that tracks readiness. Details are in his video talk.
A topic frame is our working version of the same principle. We use our own terms: the hub page answers the topic’s main question, branches solve standalone tasks, leaves cover narrow questions, and bridges lead to the next task. The audience → architecture → content order matches DrMax’s system; the terms, steps and example below are ours.
Sources: La Fabrique du Net, DrMax, Lily Ray
Frame vs keyword clusters
Clustering answers “which queries can share one page”. A frame answers a different question: “what tasks does a person have in this topic, and in what order do they solve them”. Switch the diagram at the top to see the difference on one example.
| Keyword clusters | Topic frame | |
|---|---|---|
| Unit | A group of similar queries | A user task |
| New page when | There’s a keyword group with volume | There’s a task no existing page solves |
| Duplicates | Common: “best”, “top” and “ranking” on separate pages | Ruled out: one task, one page |
| Links | Menus and “related posts” blocks | Up to the hub, down to branches, sideways to the next step |
| What a visitor sees | A set of separate articles | A path from question to decision |
Clustering doesn’t disappear: it helps you learn the query language within each task. But the decision to create a page is made by task, not by volume.
Step 1. The topic boundary
First decide what’s in the topic and what isn’t. The boundary protects against two mistakes: bloat (every adjacent query becomes a page) and dead ends (the visitor leaves for a competitor to take the next step).
- Main question. One sentence the hub page answers. In our example: “which coffee machine to get for home and how to live with it”.
- Inside the boundary: everything needed to decide, such as choosing, comparing, care and typical faults.
- Outside the boundary: neighboring topics with their own hub, such as coffee and beans, grinders, professional equipment. They get links, not pages inside the frame.
- Boundary test. If a page could move to another section without losing meaning, it doesn’t belong here.
Step 2. The audience’s voice
Tasks don’t come from guesswork or search suggestions alone. You need places where people describe the problem in their own words: forums, Reddit, marketplace reviews, support questions, comments under reviews, sales conversations.
| Channel | What to take from it | Coffee machine example |
|---|---|---|
| Support and sales questions | The tasks people come with and where they get stuck | “I want a cappuccino without fiddling with milk” |
| Marketplace reviews | The criteria people praise and complain about | Noise, warm-up time, pod price |
| Forums and Reddit | Doubts and fears before buying | “It’ll clog with scale within a year” |
| Comments under reviews | The words people use for things | “Bean-to-cup” rather than “fully automatic espresso machine” |
| Search results and “People also ask” | Questions the results don’t answer directly | “Can I use filtered water?” |
Tag each task with its source and how often it comes up. A task one person mentioned is a hypothesis. A task that repeats across sources is a branch candidate.
Step 3. Decision path and page roles
Order the tasks the way a person solves them. Each step gets the page type that handles it best.
- Understand what’s out therehub
Hub page: types, who each suits, where to go next.
- Choose a typebranch
Pods vs espresso vs bean-to-cup: criteria, limits, who each suits.
- Compare modelsbranch
A table built on criteria from the audience’s voice, not marketing specs.
- Settle doubtsleaves
Narrow questions: noise, cappuccino, budget models.
- Look after itbranch
Instructions: cleaning, scale, water. Steps, prerequisites, what to do if it fails.
- Fix itbranch
Faults: symptom, cause, what to do yourself, when to call a technician.
An illustrative example: it shows the logic, not a ready structure for a specific store.
The one-task rule: if two pages solve the same task, one is redundant. If one page solves three tasks, that’s three pages or a hub with links.
Step 4. Internal linking
Links in a frame aren’t decoration; they’re the route. We use three directions and one rule for anchors:
- Up
Every branch and leaf links to the hub page. That’s the main signal that the topic is one whole.
- Down
The hub leads to every branch, a branch to its leaves. From the text, where the question comes up.
- Sideways
A bridge to a neighboring task when it’s a real next step: from choosing a type to comparing models, from descaling to choosing water.
- Anchors
In the audience’s words from step 2. Not the same exact-match phrase on every page.
Menus, breadcrumbs and “related” blocks don’t replace contextual links. They’re useful, but the route is built inside the text.
Step 5. A brief for every page
Briefs turn the frame into work. A brief records not word count and keywords but what people will come to you for. A similar set of fields (page job, required core, replaceability) is part of DrMax’s content logic; we add the entry point and the link route, and check metadata separately at launch.
- Outcome for the reader
One sentence: what action or decision becomes possible.
- Who arrives and from where
Which branch or query, with what question.
- Must-have content
Criteria, steps, limits: what the reader would otherwise go back to search for.
- Why it beats an AI answer
Data, experience and examples that competitors and AI answers don’t have. Without that, the page isn’t ready.
- Route
Links up, down and sideways, with anchors.
- Facts and sources
What to check and where. Who owns the data that changes.
Once briefs are done and pages are written, check the frame before launch. How to do that is in part 5.
FAQ
What is a semantic cocoon?
A structure of pages around a target page, where supporting pages reinforce it and link to each other with contextual links that follow search intent. The idea was developed by French SEO Laurent Bourrelly.
How is a topic frame different from a cocoon?
It’s our working version of the same principle: our own terms (hub, branches, leaves, bridges) and a five-step process that designs pages from audience tasks rather than keyword groups.
How is a frame different from keyword clustering?
Clustering groups similar queries. A frame decides what tasks a person has and which page solves each. Clustering is used inside the frame to learn the query language.
When should a frame get a new page?
When there’s a standalone user task no existing page solves. Matching keywords alone isn’t enough.
How should links work inside a frame?
Up to the hub page, down to branches and leaves, and sideways to a neighboring task when it’s a real next step. Use the audience’s words as anchors and link from the text.
Sources
The semantic cocoon description is based on French sources on the method. DrMax’s approach is summarized from his materials. Our overall logic of audience → structure → page content is close to his system; the frame terms, step order, tables and coffee machine example are ours, and the example is illustrative.
How this piece was made: written by the SurfUp editorial team with AI assistance; every fact and number was checked by hand against the sources listed above.