The 'Sub-Intent' Protocol: Branching DOM Hierarchies and Review-Mined Categories Route Catalogs into LLM Citations

SEOs historically deployed identical programmatic review templates across broad product catalogs, but the algorithm now presumably punishes the approach with search ranking decay, and LLM engines might be even harsher.

Arnout Hellemans warns that standardizing layouts — such as placing safety and seating specs above the fold across all vehicle reviews — ignores entity-specific anxiety vectors.

Hellemans recommends configuring templates to branch by sub-intent, pulling the entity's primary concerns to the top of the DOM.

A sports car review must prioritize handling and performance metrics immediately below the header, an electric vehicle template must lead with battery range, and a minibus requires fuel economy and reliability blocks.

Hellemans advises pulling PAA questions from AlsoAsked: they are the index of what a review needs to answer, and query fan-out is very close to them.

Separately, regarding category expansion, Ross Hudgens recommends extracting hundreds of on-site reviews from product catalogs of 2,000-plus pages into a spreadsheet to mine for conversational problem entities.

Hudgens notes that Petco, for example, isolates recurring phrases like "picky eater" across dog food reviews and cross-references those patterns against Ahrefs, Semrush, or MCP search volume.

After matching those entities to search demand, one could then spin up dedicated long-tail category pages tailored to conversational AI search structures.