MAGEO framework teaches agents to reuse GEO strategies across engines
Researchers Beining Wu, Fuyou Mao, and Jiong Lin propose MAGEO, a multi-agent framework that reframes Generative Engine Optimization as a strategy learning problem, enabling reusable, engine-specific editing skills to accumulate across tasks rather than optimizing each instance in isolation.
Score breakdown
Teams building content strategies for AI-powered search engines can look to MAGEO's skill-reuse approach as a blueprint for developing transferable, engine-specific optimization workflows rather than re-solving each content task from scratch.
- 01MAGEO is a multi-agent framework that reframes GEO as a strategy learning problem, enabling reusable, engine-specific optimization skills to accumulate over time.
- 02Current GEO methods optimize each instance in isolation and cannot transfer strategies across tasks or engines.
- 03MAGEO's multi-agent system coordinates planning, editing, and fidelity-aware evaluation as its execution layer.
Beining Wu, Fuyou Mao, and Jiong Lin identify a core limitation in current Generative Engine Optimization (GEO) research: existing methods treat each content optimization task in isolation, making it impossible to accumulate or transfer effective strategies across different tasks or generative engines (GEs). Their proposed framework, MAGEO, addresses this by casting GEO as a strategy learning problem. Within MAGEO, a multi-agent system handles coordinated planning, editing, and fidelity-aware evaluation as the execution layer, while validated editing patterns are progressively distilled into reusable, engine-specific "optimization skills" that can be applied to future tasks.
They also release MSME-GEO-Bench, a multi-scenario, multi-engine benchmark grounded in real-world queries.
To enable controlled, causal assessment of content edits, the authors introduce two new evaluation contributions: the Twin Branch Evaluation Protocol, designed to attribute performance changes directly to specific content edits, and DSV-CF, a dual-axis metric that jointly measures semantic visibility and attribution accuracy. They also release MSME-GEO-Bench, a multi-scenario, multi-engine benchmark grounded in real-world queries. Ablation studies confirm that both engine-specific preference modeling and strategy reuse are central drivers of MAGEO's gains over heuristic baselines across three mainstream generative engines. Code is publicly available on GitHub.
Key facts
- 01MAGEO is a multi-agent framework that reframes GEO as a strategy learning problem, enabling reusable, engine-specific optimization skills to accumulate over time.
- 02Current GEO methods optimize each instance in isolation and cannot transfer strategies across tasks or engines.
- 03MAGEO's multi-agent system coordinates planning, editing, and fidelity-aware evaluation as its execution layer.
- 04The authors introduce a Twin Branch Evaluation Protocol for causal attribution of content edits.
- 05DSV-CF is a new dual-axis metric that unifies semantic visibility with attribution accuracy.
- 06MSME-GEO-Bench is a new multi-scenario, multi-engine benchmark grounded in real-world queries.
- 07Experiments on three mainstream engines show MAGEO substantially outperforms heuristic baselines in both visibility and citation fidelity.