6.Win the comparisons.

Aggregate numbers describe the war; questions are the battles. Once your baseline exists, the fastest way to make progress is to stop thinking about “our AI visibility” and start thinking about specific questions you are winning and losing, against specific named rivals.

Know your three question types

Discovery questionsask for options without naming anyone: the assistant’s shortlist effectively is the market, and presence there is the prize. Branded questions name you: you will almost always appear, so the game is accuracy and tone, not presence. Competitor questionsput brands side by side: “you versus them” and “alternatives to them”. These reveal who the assistant actually reaches for under pressure, and the alternatives form is a standing chance to be introduced to a rival’s dissatisfied customers.

The battle frame

For each tracked question, look at the collected answers and ask: who won this? A prompt battle has one of three outcomes. Won: you anchor the answers, named first or recommended outright. Contested: you and rivals trade position across engines and runs. Lost: someone else anchors the answers, and the scoreboard should tell you exactly who.

Naming the winner matters. “We are underperforming on comparison queries” produces meetings. “We lose this question to this rival, in these answers, because the engines keep citing these two sources where they appear and we do not” produces a task.

Picking your battles

Do not spread effort evenly. Rank your questions by commercial weight, then take them in this order: contested battles first, because you are one good source or one quotable page from tipping them. Then the lost battles with the highest stakes, worked through chapter five’s playbook: check the facts, close the source gaps the winner exploits, publish the page that answers the question better than their material does. Branded accuracy runs continuously underneath; wrong facts about you are never acceptable collateral.

Comparison answers also tell you the criteria engines think matter in your category. If every answer weighs a capability you have but never get credit for, that is not a product gap; it is a sources gap, and it is fixable with chapter five’s tools.