Skip to content
alleo wiki Homewiki
/Agentic RAG — AI Agent-Based Retrieval-Augmented Generation
📘Concept

Agentic RAG — AI Agent-Based Retrieval-Augmented Generation

최종 업데이트:

Definition

Agentic RAG is a RAG pattern in which an AI agent evaluates search results and autonomously performs additional searches and tool calls.

TL;DR

Agentic RAG goes beyond basic RAG's single search-and-generate pipeline: an AI agent explores the web in multiple steps and synthesizes information. ChatGPT's Deep Research, Perplexity's Pro Search, and Claude's agent features fall under this pattern. For content creators, a site's internal link structure and the connections between related pages become more important.

Basic RAG vs. Agentic RAG

ItemBasic RAGAgentic RAG
Number of searches1Multiple (autonomous judgment)
Tool useSearch onlySearch + calculation + API calls
Result verificationNoneAgent re-searches after evaluating results
Complex questionsSingle answerDecompose → solve step by step
Representative implementationsChatGPT BrowseDeep Research, Operator, Pro Search

How It Works

  1. User question → the agent decomposes the question into subtasks
  2. Perform a search for each subtask
  3. The agent evaluates the search results (Is it sufficient? Is it accurate?)
  4. If insufficient, perform additional searches or call other tools
  5. Synthesize all information to generate a final answer

Implications from an AEO Perspective

An agent starts from a single page and follows internal links to gather information. Therefore:

  • Sites with well-connected internal link structures have an advantage
  • Use a Pillar-Cluster structure to systematically connect related pages
  • A logical site architecture makes it easy for the agent to find related information
  • Content must be included in the HTML via SSR/SSG for the agent to read it

Frequently Asked Questions

Q. Are Agentic RAG and Multi-hop RAG the same thing?
A. They overlap but differ. Multi-hop RAG is a pattern that performs multiple search steps sequentially, whereas Agentic RAG is a broader concept that also includes tool calls and result verification beyond searching.

Related Sources

이 페이지를 참조하는 항목

관련 항목

📘ConceptPillar
Query Fan-Out
Query Fan-Out is the mechanism by which AI answer engines decompose one user question into multiple sub-queries, search many sources in parallel, and synthesize an answer.
📘Concept
GraphRAG — Knowledge Graph-Based Retrieval-Augmented Generation
GraphRAG is a next-generation RAG pattern that converts documents into a knowledge graph and then uses the graph structure to retrieve information.
📘Concept
How RAG Works
RAG is a core technology that combines retrieval and generation to improve AI answer accuracy.
📘ConceptPillar
What Is AEO?
AEO is the practice of optimizing content so AI answer engines cite it.
📘ConceptPillar
Internal Linking Strategy
Internal linking strategy is the practice of semantically connecting pages within your own site to optimize topic authority and bot and user navigation.
📙How-to
ChatGPT Citation Optimization
ChatGPT citation optimization is the work of getting content cited in ChatGPT answers.
📙How-to
Perplexity Citation Optimization
Perplexity citation optimization is the work of securing citations from a real-time web search-based AI.
📘ConceptPillar
Site Architecture
Site architecture is the overall design of page hierarchy, URL structure, and internal linking on a website. It simultaneously determines crawl efficiency, indexing quality, and user navigation experience — a foundational SEO element.

이런 항목도 있어요

📘Concept
GraphRAG — Knowledge Graph-Based Retrieval-Augmented Generation
GraphRAG is a next-generation RAG pattern that converts documents into a knowledge graph and then uses the graph structure to retrieve information.
📘Concept
How RAG Works
RAG is a core technology that combines retrieval and generation to improve AI answer accuracy.
📘ConceptPillar
LLM Optimization — An Overview of Getting Cited by AI Answer Engines
LLM optimization is a comprehensive strategy for optimizing content so that it gets cited by large language model-based answer engines such as ChatGPT, Claude, Gemini, and Perplexity.
📘ConceptPillar
What Is AEO?
AEO is the practice of optimizing content so AI answer engines cite it.
이 페이지가 도움이 됐나요?