GEO: Generative Engine Optimization — The Research Paper

Paper
GEO: Generative Engine Optimization
Submitted
November 16, 2023
Authors
Pranjal Aggarwal; Vishvak Murahari; Tanmay Rajpurohit; Ashwin Kalyan; Karthik Narasimhan; Ameet Deshpande
arXiv
2311.09735
Official source
arxiv.org/abs/2311.09735
Ivory research sheets and teal glass dividers connected by fine copper threads.
Read the methods alongside the findings to understand what was tested.

What the Researchers Studied

The paper introduces Generative Engine Optimization as a framework for improving content visibility within generative-engine responses. Its abstract describes generative engines as systems that gather and synthesize information from multiple sources, then use large language models to produce answers. It frames a visibility problem for content creators: with systems changing quickly and operating as black boxes, creators have limited control over when and how their material appears. The original paper provides the methods and full findings.

Generative Engines and Visibility

The researchers tested changes to source content, including adding citations, relevant quotations and statistics, as well as changes to fluency and presentation. These were evaluated methods, not a universal checklist or a reason to add unsupported claims.

Visibility was assessed through measures including the amount of attributed answer text and its position, alongside subjective impression measures. A source’s prominence in an answer is a different outcome from referral traffic or revenue. See the methods and evaluation sections.

GEO-Bench

GEO-Bench contains 10,000 queries assembled for evaluating generative engines. The paper’s experimental setup retrieves five search sources and generates answers using an LLM; it also reports a separate evaluation on a deployed engine. The benchmark creates a structured comparison, while the tested systems, query mix and metrics determine how to interpret the findings.

Read the experimental setup and benchmark description before applying a finding to another system.

Why the Paper Matters

The paper helped formalize the term Generative Engine Optimization and presented GEO as a framework for improving content visibility in generative-engine responses. Its abstract reports an improvement of up to 40% in visibility in the authors’ evaluation, and notes that effectiveness varied by domain. That result belongs to the study’s tested conditions; it is not a promised result for every site, query or AI system. The paper is a research milestone, not evidence that no related ideas existed before it, and it does not establish the origin of World GEO Day.

November 16 and World GEO Day

World GEO Day is observed on November 15, immediately before the anniversary of the paper submission. The connection explains the observance date; it does not suggest the authors created or endorse World GEO Day. Read the date rationale or explore the history of GEO.

Read the Original Research

Read GEO on arXiv

Related Pages

Generative Engine Optimization · history of GEO · why November 15