Habr AI shows how to build an SEO system for a niche and prepare a site for AI search

Habr AI examined why the attempt to "hand SEO over to a neural network" most often ends in noise, duplicates and hallucinations. The article's main point: winning a niche starts not with text generation, but with demand engineering and a precise site architecture.

Demand first

The author pushes back on a popular idea that LLMs can assemble the semantic core on their own, invent the structure and fill the site with content. In practice, this approach yields hundreds of attractive but useless queries, pages with overlapping meaning, and headings that don't answer users' real intents. The problem isn't the model itself, but that it is often handed chaos without rules. If a business doesn't understand what kinds of demand exist in the niche and which pages should cover them, automation only scales the errors.

Instead of an abstract "make it good," the author suggests starting with a demand map: which queries belong to commercial search, which to solution comparison, which to navigation, and which don't need a separate page at all. This changes the logic of the work: the site is built not around a set of keywords, but around intent types and user scenarios. That makes it clear where a separate URL is needed and where it is enough to strengthen an existing page.

"Chaos doesn't become a system just because you added an API key to it."

A pipeline without magic

The article describes a fairly down-to-earth but workable pipeline. First, the team collects raw queries and cleans out duplicates and junk. Then clusters are tied to specific page types: where a commercial landing page is needed, where a catalog filter, where a review, where a comparison page, and where it's better not to publish at all. Only after that can an LLM be plugged in, not as a replacement for strategy, but as a tool to speed up analysis, drafts and the scaling of a structure that has already been decided.

• Raw queries and initial cleanup
• Clustering by intent, not just by words
• Mapping clusters to page types
• A plan for development, content and internal linking
• Preparing pages for visibility in AI search

The key point is that the semantic core stops being a "graveyard of spreadsheets." It becomes a production process in which every cluster has an owner, a page format, a set of requirements and a clear priority. This approach is useful not only for the SEO team but also for product, development and editorial: everyone can see which pages the business needs, which can be merged, and which are better left unlaunched so as not to dilute the site's structure.

Where LLMs are useful

In this scheme, LLMs don't cancel SEO; they take a specific place inside the pipeline. They can help with normalizing phrasings, initial query grouping, generating structure options, drafting metadata and analyzing gaps in niche coverage. But the model must work to set rules and on verifiable data. Otherwise you get a familiar picture: hundreds of duplicate pages, cannibalization of search demand, and texts that look convincing but don't solve the user's problem and don't strengthen the site as a system.

The shift toward AI search matters separately. Where it used to be enough to think only in terms of classic search results, now you have to consider how material will be read and extracted by systems such as AI Overviews and other LLM interfaces. That calls for a transparent structure, a clear match to intent, no duplication, and page logic that a machine can interpret without guessing. Otherwise even a quality text ends up in a weak position, because no clear information model is built around it.

What this means

For editorial teams, marketing and product teams, the takeaway is simple: the LLM era doesn't cancel manual demand design, it makes it even more important. The winners won't be those who publish thousands of AI texts fastest, but those who understand intents better, build architecture more carefully and use models as an automation layer on top of a well-thought-out system. That discipline is what raises the odds of appearing not only in ordinary search results, but also in AI answers.