IntentGEO: Intent-aware Generative Engine Optimization for Recommendation
| Source: arXiv AI
Tags: Generative Engine Optimization, IntentGEO, AMO, Doubao, LLM recommendation
IntentGEO shows that generative engine optimization gains depend on user intent, and its Anchored Meta-Optimization method raised average citation-rank gain by 35% over the best prior learning-based baseline, including on the commercial assistant Doubao.
Details
Generative Engine Optimization (GEO) aims to make items more visible in LLM-generated recommendations, but most work optimizes content for generic queries. IntentGEO evaluates shared rewrites across different need specificity and decision stages. The authors find that existing GEO gains vary by intent and that learning-based prompt optimization can underperform its handwritten seed, drifting from recommendation instructions toward generic promotional tactics. Their fix, AMO (Anchored Meta-Optimization), keeps recommendation instructions in a fixed core and learns appended strategies from citation feedback. Across four domains and ten intent states, AMO improved average citation-rank gain by 35% over the strongest prior learning-based baseline, led on nine states, and also scored highest in an agentic pipeline and, after adaptation, on Doubao.