Perplexity launches search API for AI

Perplexity has launched a search API that lets developers connect their applications to the web index used by its own search engine. The company says it covers hundreds of billions of web pages and is designed to power AI systems, agents, and traditional applications.

The tool was introduced on September 25, 2025, alongside an SDK, an open-source evaluation framework, and a new platform for managing the API and its documentation.

Search ready to use in AI applications

Until now, building an application capable of searching the web for up-to-date information required combining several services: web crawling, storage, document processing, and a system for finding relevant passages. Perplexity wants to reduce that work to a single interface.

The API does not return only complete pages. It breaks documents into small units, compares them with the query, and delivers the most useful passages already ranked. In practice, an assistant that needs to answer questions about a recent news story, scientific research, or a product can receive the most relevant context directly, without having to process the entire document from scratch.

That could mean faster integrations and less in-house infrastructure for development teams.

The goal: reduce outdated information

One common problem with AI agents is working with old information. A model can write a convincing answer, but fail if its data does not reflect what happened a few hours ago.

Perplexity says its systems process tens of thousands of index update requests every second. It also says it receives millions of user queries every hour, signals it uses to continuously improve how it interprets and organizes web content.

The company also uses an AI-based content understanding module. Its purpose is to adapt to the variety of formats found online, from well-structured articles to pages that are difficult to interpret, and extract the information that may be useful for a search.

These figures and capabilities are claims made by Perplexity, not an independent audit of its performance.

A new option for developers

The company says its API offers a competitive combination of relevance, speed, and cost. To test it, Perplexity has published search_evals, an open-source framework that lets developers compare different search APIs on direct search tasks and more complex research workflows.

It has also introduced an SDK, meaning a set of tools that makes it easier to integrate the service into an application. According to Perplexity, its own engineers have built prototypes in less than an hour by combining it with AI-assisted programming tools.

For you, the change is not in the search engine you use today, but in what may appear inside other applications: assistants that can look up recent information, systems that prepare reports with sources, or products that answer questions about constantly changing data.

The API does not remove the need to verify answers. Finding up-to-date information does not guarantee that a system will interpret it correctly or that every source will be reliable. The important questions will be how it performs outside Perplexity's tests, what prices it offers at scale, and how much accuracy it maintains when developers start using it in real products.