Category Teardown

What it produces

A category teardown: a structured document showing who owns visibility in your category inside AI search, what the engines say about each player, where the whitespace sits, and the three moves that would shift your position. Input is your category plus two competitors. Runtime is roughly 45 minutes.

INPUT        PERPLEXITY      CLAUDE        NOTEBOOKLM     OUTPUT
category  ▶  scout the   ▶  synthesise ▶ interrogate ▶ teardown
+ 2 rivals   citations      + score       the findings   + 3 moves

The pipeline

1. Stage 1 — Perplexity (scout)

   Run the category's top buyer questions. Capture which brands get cited, in what order, and with what framing. Perplexity shows its sources, so you read the citation graph directly.

2. Stage 2 — Claude (synthesise)

   Feed the raw citations to Claude. Ask it to cluster the framing, score each competitor's visibility, and name the gaps. This is where the teardown takes shape.

3. Stage 3 — NotebookLM (interrogate)

   Load the synthesis plus source pages into NotebookLM. Pressure-test the findings — "where is the consensus weakest," "which claim is least defended" — before you act on them.

STAGE 1  "List the top 8 questions a buyer asks
          before choosing a <category> brand."
STAGE 2  "Here are the cited sources per question.
          Cluster the framing, score each brand 0-3
          on visibility, name the three gaps."
STAGE 3  "Challenge this synthesis. Where is the
          weakest claim, and what would disprove it?"

Ship the outcome

The output is a decision input, not a report to file. The three moves at the end are what you carry into the next planning session. Run it monthly per category to track how your AI-search position moves.