Agentic search
AI agents that autonomously search for information, deciding what and how to look.
Agentic search is a way for AI to find information. Instead of just waiting for a user to ask a specific question, an AI agent can figure out what it needs to know and then go search for it. It's like having a smart assistant that can do its own research.
For example, imagine you ask an AI to plan a trip. An agentic search system might decide it needs to find flight prices, hotel availability, and local attractions. It would then perform separate searches for each of these things, gather the information, and put it together for you.
Agentic search involves AI agents that autonomously plan and execute information retrieval tasks. The agent interprets a goal or query, breaks it down into sub-tasks, selects appropriate search tools or methods, and synthesizes the results. This contrasts with traditional search where a user explicitly defines the query. Tradeoffs include increased autonomy and potential for novel information discovery versus the risk of inefficient or misdirected searches if the agent's planning is flawed.
Agentic search is gaining attention as AI models become more capable of understanding complex user needs and autonomously navigating information sources. Recent discussions highlight its role in improving search functionality and enabling AI agents to perform more sophisticated information retrieval tasks.
- An alignment assessment of recent cybersecurity incidentsAnthropic · Anthropic Research ·
- Anthropic Cookbook: Restore the Tools category on the agentic-search registry entry (#810)Anthropic · Anthropic Cookbook ·
- Three Kinds of Software Survive: Tasklet's Andrew Lee on Competing to be a Horizontal PlatformAnthropic · The Cognitive Revolution ·
- Anthropic Claude
- Anthropic is actively developing and discussing agentic search, as seen in their efforts to restore categories on an agentic search registry and their focus on AI alignment in relation to model access.
- Web search
- Traditional web search requires explicit user queries, whereas agent
- Question answering
- Question answering systems typically provide direct answers
- Information extraction
- Information extraction focuses on pulling specific data
The concept of agents autonomously searching and acting has roots in AI research on intelligent agents and planning. The specific term 'agentic search' is emerging in the context of large language models being equipped with tools and the ability to reason about their use.
Ask a chatbot like ChatGPT or Claude to 'plan a weekend trip to San Francisco for two people.' Observe if it asks clarifying questions or if it directly starts searching for flights, hotels, or activities on its own. This can show how it might autonomously decide what information is needed.
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- ChecklistCyber evaluation incident transcript review processSteps to scan and identify internet access incidents in model transcripts. · 30 min
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
Questions people ask
- How is agentic search different from regular search?
- Regular search relies on you telling the system exactly what to look for. Agentic search lets an AI figure out what information it needs and then go find it by itself.
- Can agentic search make mistakes?
- Yes, because the AI is making decisions, it can sometimes search for the wrong things or get stuck in loops if not properly guided or if its planning capabilities are limited.
First seen on Subvolts Aug 12, 2026. Updated Oct 19, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.