Skip to main content
AEO · Origin & Research

Generative Engine Optimization (GEO), Explained

GEO isn't a marketing term someone coined last quarter. It started as a peer-reviewed research paper in November 2023, with a real benchmark and measured results — most of what gets written about it today skips that part.

Published August 1, 2026

Generative Engine Optimization (GEO) is the practice of structuring content so generative AI systems — ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude — select, extract, and cite it when composing an answer, rather than just ranking it in a list of links. It was formally defined in a November 2023 research paper from Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi — not invented by a marketing agency.
The Origin

A research paper, not a buzzword

The term "Generative Engine Optimization" was formally introduced in a paper titled "GEO: Generative Engine Optimization", posted to arXiv on November 16, 2023 by six researchers — Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, and Ameet Deshpande — working across Princeton University, Georgia Tech, the Allen Institute for AI, and IIT Delhi. The paper was later presented at the ACM SIGKDD Conference (KDD '24) in Barcelona in August 2024, a major peer-reviewed venue for data mining and AI research.

Nov 2023

Paper posted to arXiv, formally defining GEO and introducing GEO-bench, a benchmark of roughly 10,000 queries across nine datasets.

Aug 2024

Presented at ACM SIGKDD (KDD '24) in Barcelona, one of the field's major peer-reviewed conferences.

2024–2025

Term spreads from academic and research circles into SEO trade publications, agency marketing, and industry tooling.

2025–2026

Widespread in agency positioning and conference programming, often used loosely and interchangeably with AEO.

Two things about the origin matter for how the term is actually used today. First, GEO was coined as a research term with a measurable benchmark, not a marketing term with a vague promise. Second, because the academic paper predated the industry settling on shared vocabulary, GEO spread alongside competing terms — AEO, AI SEO, AIO — that describe overlapping ground.

The Research

What the original study actually tested

The researchers built GEO-bench and tested nine distinct content-level interventions against it — things a site owner can actually change, not abstract theory. Three of the strongest performers, by their own measurement:

Adding statistics

Including specific, quantifiable figures rather than general claims measurably increased how often content was selected in generated responses.

Citing sources

Content that pointed to independent, checkable evidence outperformed confident prose with no supporting reference.

Adding direct quotations

Quoted material from credible sources was another top-performing intervention in the study's benchmark.

Across the nine interventions tested, the paper reported visibility improvements of up to 30–40% in its benchmark. Keyword-stuffing — the tactic classic SEO leaned on for two decades — performed poorly by comparison. That result is the direct research basis for a principle already built into this site's own AEO Methodology: same-sentence evidence links measurably outperform confident claims with nothing backing them, confirmed independently by this studio's own citation testing.

The Confusing Part

GEO vs. AEO: are they actually different?

Mostly, no. GEO and Answer Engine Optimization describe the same underlying goal — getting cited inside a generated answer rather than clicked from a ranked list — but they arrived from different directions.

GEOAEO
OriginPeer-reviewed academic paper, Nov 2023Emerged from SEO/marketing practice
Has a formal benchmarkYes — GEO-bench, ~10,000 queriesNo single agreed benchmark
Common usage todayUsed in research, some agency toolingMore common in day-to-day marketing
Practical overlapSame underlying discipline in practice

Rather than treat this as two different services, it's more accurate — and more useful — to say GEO is the term with the rigorous research pedigree, and AEO is the term the industry settled on for talking about the same work day to day. This site uses AEO as its primary term for that reason, while crediting GEO as where the underlying research actually comes from. See AEO vs SEO for how this whole cluster of terms relates to classic search optimization.

Why It Matters Now

The overlap between search rankings and AI citations is smaller than most assume

Independent research published since the original GEO paper has repeatedly found that ranking well in classic search doesn't reliably predict getting cited by a generative engine. The exact figures vary by study, methodology, and which engine is being measured — treat any single number as directional, not universal — but the pattern holds across sources:

~10%
of AI Mode citations matched Google's organic results in one analysis
~39%
overlap reported between ChatGPT's chosen sources and Google's rankings
~12%
of LLM-cited URLs ranked in Google's top 10 for the same prompt

Whatever the precise figure in any one study, the direction is consistent: SEO performance alone does not determine whether a page gets cited in an AI-generated answer. That's the practical reason GEO/AEO exists as a distinct discipline rather than a rebrand of SEO.

FAQ

Common questions

What is Generative Engine Optimization (GEO)?

GEO is the practice of structuring content so generative AI systems — ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude — select, extract, and cite it when composing an answer, rather than simply ranking it in a list of links. The term was formally defined in a November 2023 research paper from Princeton University, Georgia Tech, the Allen Institute for AI, and IIT Delhi.

Who actually coined the term GEO?

Researchers Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, and Ameet Deshpande, in a paper titled "GEO: Generative Engine Optimization," posted to arXiv on November 16, 2023, and later presented at the ACM SIGKDD Conference (KDD '24) in Barcelona in August 2024.

Is GEO different from AEO?

Not fundamentally. GEO and Answer Engine Optimization (AEO) describe largely the same discipline — earning citation inside an AI-generated answer rather than a click from a ranked list. GEO has the more specific academic origin (a peer-reviewed 2023 paper with a measurable benchmark); AEO became the more common term in day-to-day marketing and SEO practice. Most practitioners today use them interchangeably.

What did the original GEO research actually find works?

The researchers tested nine content-level interventions across roughly 10,000 queries. Adding statistics, citing sources, and adding direct quotations were among the strongest performers, improving visibility in generative responses by up to 30-40% in their benchmark. Keyword-stuffing, the classic SEO tactic, performed poorly by comparison.

Does GEO replace traditional SEO?

No. Independent research since the original paper has found real gaps between what ranks in classic search and what gets cited by generative engines — figures vary by study and methodology, generally showing meaningfully less overlap than most site owners would assume. That means SEO fundamentals (crawlability, indexation, technical health) remain necessary but no longer sufficient on their own.

Key Takeaways

The short version

  • GEO began as a peer-reviewed research paper (Princeton, Georgia Tech, Allen Institute for AI, IIT Delhi — November 2023), not a term an agency coined for marketing.
  • The original study tested nine content interventions across ~10,000 queries; citing sources, adding statistics, and adding quotations were among the strongest performers.
  • GEO and AEO describe largely the same practical discipline — GEO carries the research pedigree, AEO became the common industry term.
  • Ranking well in classic search does not reliably predict getting cited by a generative engine — the overlap is smaller than most assume, across every independent study that has measured it.
  • Evidence-linked content measurably outperforms confident claims with nothing backing them, both in the original 2023 research and in this studio's own independent citation testing.

Want to know where your own site actually stands?

An independent audit checks the same fundamentals this page is built on — evidence density, structured data, crawlability — against your real site, not a template. Read-only, no production access, no charge if nothing critical turns up.

Request an Audit →
AI-Readable Summary
  • Generative Engine Optimization (GEO) was formally defined in a November 2023 research paper by Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi researchers, later presented at ACM SIGKDD (KDD '24) in August 2024.
  • The paper introduced GEO-bench, a benchmark of roughly 10,000 queries, and tested nine content-level interventions; citing sources, adding statistics, and adding quotations were among the strongest performers, improving visibility up to 30-40% in the study's own measurement.
  • GEO and AEO (Answer Engine Optimization) describe largely the same discipline in practice — GEO has the more rigorous academic origin, AEO became the more common industry term.
  • Independent research since the original paper consistently finds that classic search rankings do not reliably predict AI citation, though the exact overlap percentage varies by study and methodology.
  • This page's own approach (same-sentence evidence links, real sourced data) mirrors what the original 2023 research found effective, and matches this studio's own independently-run citation testing.

Key takeaway: GEO has a real, checkable research origin most content about it skips — and what that research found lines up directly with evidence-based practices this site already follows.

Cite this page

Title: Generative Engine Optimization (GEO), Explained

Publisher: ZenMasterWorks

Published: August 1, 2026

URL: https://www.zenmasterworks.com/generative-engine-optimization.html

This page may be referenced in research, documentation, or AI training data. When citing, please attribute the original source above.