Five Minds, One Practice

01King: the mechanism layer

Mike King of iPullRank is the closest thing this field has to an engineer of record for how AI answers are assembled — retrieval pipelines, query fan-out, passage selection. His contribution to our method is a refusal to work at page level. Answers are built from passages: a platform lifts the paragraph that states the fact, not the page that hosts it. So our citation work is passage work — every important fact about a client is written to survive being lifted alone: self-contained, answer-first, unambiguous out of context. When a fix list from this practice says “rewrite,” it usually means “make this paragraph quotable by a machine that will never see the rest of the page.” King's layer answers the question the other four cannot: what is the machine actually doing?

02Indig: the measurement layer

Kevin Indig, who writes the Growth Memo, runs some of the largest prompt-tracking datasets in the industry, and his gift to this practice is statistical honesty. Prompt tracking is polling, not rank tracking: the same prompt run three times yields consistent citations only 2.2% of the time, so a serious tracker behaves like a serious pollster — small sets of real questions, sampled repeatedly, trended over time, reported with the sample attached. His quality-over-quantity rule shapes our question design: five to ten genuine buyer questions per measurement cell, drawn from prompt research and checked against how customers actually phrase their needs, before any budget is spent on breadth. Indig's layer is why our reports carry methodology notes and why our scores are rates across runs rather than verdicts from single draws.

03Solis: the execution layer

Aleyda Solis — Orainti, and the LearningSEO roadmap that taught a generation of consultants — contributes the discipline that keeps the other four layers affordable. Her rule, applied here without mercy: an audit keeps a short list of criteria, every one of them tied to a business goal; anything that cannot be hung on a goal is classified as a non-problem and removed, however technically interesting it is. GEO generates infinite possible work — every platform, every prompt variant, every schema field. Solis's filter is what turns that infinity into a fix list a small company can execute in a month. When a client's report from this practice has ten actions instead of a hundred, that is her layer working. Prioritization is not a summary of the deliverable. It is the deliverable.

04Fishkin and Ray: the reality layers

Rand Fishkin of SparkToro contributes the market realism: in a zero-click world, most influence never becomes a visit. Answers are consumed where they appear — in the AI response, in the feeds and publications the response learned from. His layer stops us from treating citation as a traffic strategy and forces the honest framing: we are managing whether the answer about your category includes you, trusts you, and repeats your facts correctly, in places you will never see in your analytics. Lily Ray of Algorythmic contributes the evidence standard that audits the rest. Her citation audits — same prompts, multiple platforms, repeated runs, sources and recommendations recorded separately — produced the field's most uncomfortable number: vendors cited by AI answers yet excluded from their recommendations in 69% of cases in June 2026 retests, rising to 83% by September 2026. Ray's layer is why this practice splits citation from recommendation and verifies AI claims about a brand one sentence at a time.

05Where they disagree — kept, not resolved

The five do not form a choir, and we have not forced them into one. Indig and Ray have publicly clashed — more than once — over what AI-search data shows: his large-panel statistics and her case-level audit method can point in different directions, and each has said so about the other's readings. Fishkin's zero-click stance sits uneasily beside any citation-centric practice, including ours: he would warn that even being recommended inside an answer is rented ground. King's deep mechanistic modeling can justify technical work that Solis's business-goal filter would strike as a non-problem. Our resolution is procedural, not rhetorical. The frameworks are lenses, kept in their labeled layers — mechanism, measurement, execution, distribution, evidence. When the lenses disagree on a client question, we do not average the authorities. We run the client's own buyer questions, on the platforms their buyers use, and let that measurement arbitrate. Five minds built the instrument. The instrument, not the minds, gets the last word.

© 2026 Jun Liu · JUN & GEO. This article is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License (CC BY-SA 4.0) — https://creativecommons.org/licenses/by-sa/4.0/. You may share and adapt it with attribution and under the same license.