AI Agents
AI Agents are autonomous systems that use LLMs to perceive their environment, make decisions, and take actions to accomplish goals. They can use tools, run code, and execute multi-step workflows without human intervention.
Definition
AI agents extend beyond single-turn generation by maintaining goals, planning sequences of actions, and adapting based on outcomes. An agent can search the web, read pages, extract information, make decisions, and take actions toward an objective. Function calling enables agents to use tools like search engines, calculators, or APIs.
Agents represent a shift from passive AI systems that respond to queries to active systems that accomplish tasks. In AI search contexts, agents can dive deeper into topics, compare multiple sources, and synthesize more nuanced answers by actively researching rather than passively citing retrieved documents.
Why it matters for AI visibility
Agentic AI search systems actively research answers rather than passively retrieving sources, potentially looking across your entire website, checking claims against sources, and verifying information. Your brand's credibility becomes more important: agents validate claims and penalize sources that contradict other information.
Related terms
Function Calling (Tool Use)
Function Calling is the capability of LLMs to call external functions or APIs as part of generating responses. It enables agents to take actions, query databases, or access real-time information beyond the model's training.
AIAgentic Search
Agentic Search refers to search and answer engines powered by AI agents that actively research topics by performing multiple searches, visiting pages, and synthesizing information autonomously.
AIAgentic Workflows
Agentic Workflows are automated processes orchestrated by AI agents that execute multi-step tasks, make decisions between steps, and adapt to outcomes. They combine language understanding with action execution.
AIAI Inference
AI Inference is the process of running a trained model on new input to generate output. It is the runtime execution phase, as opposed to training, where the model learns patterns from data.