Find the gaps: questions, competitors and sources.

Start from the answers, not the score. For every buyer question you track, note whether you were named, who was named instead, and which pages the engine cited. A low score is many different problems added together; this list separates them.

Then sort each lost question into one of two piles. In the first, someone else was named: there is a shortlist to get onto, and the cited pages show where it is assembled. In the second, nobody was named: the engine answered with criteria or prices, and the gap is a guide page that explains how to choose. In the public Tesla example, two of five ways of asking about grid storage named no supplier in any of 24 answers (the phrasing study).

For “best tool for X” questions, engines lean on pages that compare options: listicles, alternatives pages, versus pages, review sites, videos and forums. In the Tesla example, youtube.com was cited 29 times, en.wikipedia.org 24 times and reddit.com 20 times, against 17 for tesla.com. Most of these pages are written by third parties, and many accept submissions or corrections.

The global most-cited lists are a poor guide to your own. Ahrefs' September 2026 list for ChatGPT puts Reddit first, but it pools every kind of question, from recipes to symptoms. Across 15 categories of buyer questions, 94% of the 4,190 domains cited appeared in one category only, and in 8 of the 15 categories none of the seven universal sites made the top ten (your category has its own cited sites). Build your list from your own questions:

1. Ask each engine the buyer questions that matter most to your pipeline.
2. List every page they cite, and mark which ones mention you, and whether they describe you correctly.
3. Sort the ones that leave you out by how often they are cited. Those are your targets.

Citation share is the term for how much of that list points at you. Working the list is chapter 5.