How AI engines pick which brands they cite.

Plain-English explanations of how AI answer engines surface, cite, and recommend brands. Written from the audits we’ve run, and free to read.

Stylized illustration of an AI control panel
Cornerstone reads Start here
Go deeper 9 more
11

ChatGPT is half the market now. We checked how often the four engines agree.

Comscore shows ChatGPT falling to 50% of AI prompt volume. We asked four engines 260 buying questions: they split on 34%, and agreed on the top brand only 23% of the time.

Original data 6 min
12

“Models aren’t deterministic.” We asked the same questions three times to see how much.

Whether a business appeared changed in 10% of repeated asks. Which brand came first changed in 46%. What that means for AI rank tracking.

Original data 6 min
10

Google says you don’t need to optimize for AI. We agree — about three of the four things they listed.

Google’s June 15, 2026 AI optimization guide rejects llms.txt, special schema, content chunking, and AI-specific writing. In our audits, three of those mythbusters check out. One doesn’t.

Strategic 8 min
04

The fall of blue links and the rise of GEO.

Zero-click search passed 50% by 2019. Gartner forecasts a 25% drop in classical search volume by 2026. Why the discipline shifted from ranks to citations.

Foundational 7 min
05

What the 2024 Google API leak actually told us.

14,014 attributes across 2,596 modules. siteAuthority is real. NavBoost tracks click quality. Topical cohesion matters.

Technical 8 min
06

Relevance Engineering: the discipline replacing classical SEO.

Mike King’s framing: the unit of work moves from the page to the passage. Vector retrieval, BM25 fusion, and the role 95% of SEO software cannot serve.

Strategic 8 min
07

How AI Mode actually works: fan-out, dense retrieval, pairwise ranking.

Six Google patents read together describe one architecture: one query becomes nine, passages embed as vectors, a language model picks winners pairwise.

Technical 9 min
08

Technical foundations: the five pillars that decide retrievability.

Site information architecture, entity mapping with JSON-LD @graph, crawlability, rendering strategy, performance: the bar before content competes.

Technical 8 min
09

GEO measurement frameworks: the metrics that actually matter.

iPullRank named the Measurement Chasm. Five things to measure across it, and why one AI answer is only a sample.

Strategic 6 min
Want a real audit?

The guide explains the method. The audit applies it.

See what ChatGPT and Google’s AI tell buyers in your city about your business. The $20 Snapshot asks about 25 of their questions; the $200 Full Audit asks about 100.