09-geo-generative-engine-optimization

Installation
SKILL.md

Generative Engine Optimization (GEO)

Apply ../00-blog-writing-guardrails/SKILL.md first; its originality, accuracy, and E-E-A-T gates are hard requirements for any post, independent of whether it also targets answer-engine citation. The gates below are additive, not a substitute.

Generative Engine Optimization (GEO) originates from a controlled black-box experiment showing that specific content interventions can boost a source's visibility in generative-engine responses by up to 40%, with efficacy varying by domain [ARXIV-GEOFOUND-01]. The gates below operationalize the interventions that follow-on research has since tested and qualified.

0. Evidence-quality caveat (read before applying any gate below)

  • GEO is an early-stage research area. Across the published literature, "terminology, metrics, and evidence standards remain heterogeneous" [ARXIV-GEOSURVEY-01], and controlled benchmarks found that general heuristics "generalize poorly," with only three of fifty-four method-domain combinations tested showing a significant positive effect [ARXIV-GEOSURVEY-01]. Treat every tactic below as a directionally supported hypothesis, not a guaranteed outcome, and re-run your own before/after comparison rather than assuming a cited percentage transfers to your domain or model mix.
  • Citation behavior varies substantially by model: one controlled study found Kimi-K2 sensitive to 83% of tested content factors versus 50% for Claude-3.5 and 33% for Gemini-2.5 [ARXIV-GEOCITE-01]. Do not optimize for a single answer engine and assume the result generalizes to all of them.
  • Do not present any GEO tactic to a stakeholder as a guaranteed ranking, citation, or traffic outcome. State it as a research-backed direction with cited, dated evidence.

1. Passage self-containment gate

  • Write the passage most likely to answer the post's core question as a direct, self-contained statement that does not depend on surrounding sentences for its subject, verb, or referents; an answer engine extracts and quotes passages independent of their context. [GOOGLE-SNIPPET-01]
  • In controlled testing, topical relevance and a source's position within the retrieved context were "the two most robust factors" in whether a generative engine cited it, with position "equally consequential" to relevance [ARXIV-GEOSURVEY-01]. Place the self-contained answer near the top of the section that addresses it, not buried after preamble.
  • Self-containment gains measured in these studies assumed the source was already retrieved into a fixed context window; they say nothing about whether the passage gets discovered or ranked in the first place [ARXIV-GEOSURVEY-01]. Pair this gate with normal SEO crawlability and indexing practice — it does not replace it.

2. Statistic and citation density gate

Installs
1
GitHub Stars
1
First Seen
Aug 12, 2026
09-geo-generative-engine-optimization — wakqasahmed/blog-writing-skills