GEO Research
The peer-reviewed academic evidence base behind the LAVF framework. All papers are peer-reviewed or preprint research on Generative Engine Optimization and AI citation behavior. Each paper links directly to its arXiv or ACM source.
Foundational Paper
GEO: Generative Engine Optimization
Pranjal Aggarwal, Vishvak Murahari, Tanmay Rajpurohit, Ashwin Kalyan, Karthik Narasimhan, Ameet Deshpande
Generative engines answer queries by synthesising information from multiple sources and summarising it with an LLM, leaving content creators little control over when and how their content appears. This paper introduces Generative Engine Optimization (GEO), a black-box optimisation framework for improving content visibility in generative engine responses, together with GEO-bench for systematic evaluation. Our summary of the authors' abstract; see arXiv for the original.
Key Findings
- →Introduces GEO — the first framework for optimising content visibility in generative engine responses
- →Reports that GEO can boost visibility by up to 40% in generative engine responses
- →Introduces GEO-bench, a large-scale benchmark of diverse user queries across multiple domains
- →Finds that strategy effectiveness varies across domains, indicating a need for domain-specific methods
Additional Research(6 papers)
Generative Engine Optimization: How to Dominate AI Search
Mahe Chen, Xiaoxuan Wang, Kaiwen Chen, Nick Koudas
Large-scale controlled comparison of AI search and traditional web search across verticals, languages and query paraphrases. Reports a systematic bias in AI search toward earned media — third-party authoritative sources — over brand-owned and social content.
Source Coverage and Citation Bias in LLM-based vs. Traditional Search Engines
Peixian Zhang, Qiming Ye, Zifan Peng, Kiran Garimella, Gareth Tyson
Empirical study of 55,936 queries across six LLM-based search engines and two traditional engines. Reports that LLM engines cite a more diverse set of domains — 37% of domains unique to them — but do not outperform traditional engines on credibility, political neutrality or safety.
Diagnosing and Repairing Citation Failures in Generative Engine Optimization
Zhihua Tian, Yuhan Chen, Yao Tang, Jian Liu, Ruoxi Jia
Proposes a taxonomy of citation failure modes across the citation pipeline, plus an agentic system that diagnoses a failure and applies a targeted repair. Reports over 40% relative improvement in citation rates while modifying about 5% of content, against 25% for baselines.
AgenticGEO: A Self-Evolving Agentic System for Generative Engine Optimization
Jiaqi Yuan, Jialu Wang, Zihan Wang, Qingyun Sun, Ruijie Wang, Jianxin Li
Frames GEO as a content-conditioned control problem and evolves a diverse archive of strategies instead of applying fixed heuristics. Reports outperforming 14 baselines across 3 datasets on two generative engines.
Structural Feature Engineering for Generative Engine Optimization: How Content Structure Shapes Citation Behavior
Junwei Yu, Mufeng Yang, Yepeng Ding, Hiroyuki Sato
Decomposes content structure into macro (document architecture), meso (information chunking) and micro (visual emphasis) levels. Across six generative engines, reports a 17.3% improvement in citation rate and 18.5% in subjective quality.
Think Before Writing: Feature-Level Multi-Objective Optimization for Generative Citation Visibility
Zikang Liu, Peilan Xu
Optimises over interpretable structural, content and linguistic features rather than rewriting text token by token. Reports that citation behaviour is influenced more by document-level content properties than by isolated lexical edits.
How we apply this research
See how the LAVF framework operationalises these findings into measurable AI visibility scores.
Papers are linked directly to their original arXiv or ACM sources. LLMs Network does not host or redistribute paper content. TL;DR summaries are our own interpretations for practitioner audiences.