Research Publications Methodology About Team Contact
Generative Intelligence Research

The Scientific Institute for Generative Intelligence

Original research into how AI systems choose what to cite. We study large language model source selection, citation behaviour, and the emerging science of Generative Engine Optimization.

82%
LLM Citation Accuracy Variance
(Industry research, 2025)
41%
Source Selection Overlap Between Models
(Princeton University, KDD 2024)
96%
Structured Content Preference Rate
(Industry research, 2025)
6.5×
Citation Lift From GEO Optimisation
(Industry research, 2025)
Research Focus

What We Study

Core Research
LLM Citation Behaviour

Investigating how large language models select, rank, and present sources in generated responses across platforms including ChatGPT, Gemini, Perplexity, and Claude.

Pattern Analysis
Generative Search Patterns

Mapping the emerging patterns in how AI-powered search engines retrieve, synthesise, and attribute information differently from traditional search.

Applied Science
Content Architecture for AI

Studying how content structure, semantic clarity, and technical implementation influence whether AI systems surface and cite a given source.

Methodology
GEO Methodology

Developing rigorous, reproducible methodologies for Generative Engine Optimization research that meet rigorous research standards of evidence and transparency.

Latest Publications

Recent Research

March 2026
Citation Patterns in Multi-Model Generative Search
J. Tavitian, V. Tavitian

An analysis of how different LLM platforms select and present sources when answering identical queries, revealing significant variance in citation behaviour.

LLM Citations Multi-Platform
February 2026
Structural Determinants of AI Source Selection
J. Tavitian

Examining which content architecture patterns most consistently correlate with inclusion in LLM-generated responses across commercial AI platforms.

Content Architecture GEO
January 2026
Toward a Reproducible GEO Research Framework
V. Tavitian, J. Tavitian

Proposing standardised methodology for studying generative engine optimization, including query design protocols and cross-platform testing frameworks.

Methodology Framework

Access Our Research

Our publications are freely available. Explore our research into LLM citation behaviour and generative engine optimization.