In short. Brave Search API gives applications access to an independent web index through search results, extracted context, and sourced answers.

Definition: Brave Search API is an application programming interface that supplies applications with data from Brave Search's independent web index. As of 17 October 2026, Brave documents services for ranked search results, extracted page content for AI models, and answers supported by web sources.

Key facts

Brave's Search API overview identifies the underlying index as the same one that powers Brave Search, developed independently of Google and Bing.
The Web Search documentation describes an option for up to 5 additional excerpts per search result.
Brave's LLM Context documentation describes extracted page passages and source metadata for large language models, or LLMs.
The Answers documentation describes a separate service that produces answers with source attribution.

What does a Web Search result contain?

Brave's API overview describes Web Search output as ranked titles, URLs, and short snippets, with local and rich result enrichments. A snippet is an excerpt displayed with a search result.

The Web Search guide shows this request for additional excerpts:

curl "https://api.search.brave.com/res/v1/web/search?q=python+web+frameworks&extra_snippets=true" \
  -H "X-Subscription-Token: <YOUR_API_KEY>"

The example searches for Python web frameworks. The API key goes in the X-Subscription-Token header. With extra_snippets=true, entries in the response's web.results array include an extra_snippets array containing alternative excerpts. The same guide documents country targeting and filtering by content language, allowing an application to specify the market and language of its search.

How does LLM Context supply material to an AI application?

Brave's LLM Context guide describes a retrieval service for agents and chatbots. It returns query-relevant chunks, meaning portions of page content, ready for a model to process without a further scraping step. Grounding is the use of retrieved material to support a model's response.

The response separates extracted content under grounding from metadata under sources. In grounding.generic, each result includes a URL, title, and snippets array. Source metadata is keyed by URL, preserving the connection between a passage and its originating page.

An application can set maximum_number_of_tokens to control the approximate amount of context and maximum_number_of_urls to limit source URLs. Tokens are the units used to measure model input. The context_threshold_mode setting controls relevance filtering. Goggles, Brave's custom source ranking rules, allow source restrictions or exclusions. These settings shape the material supplied to the application's own model.

What does this change for a website?

Because the API uses Brave Search's index, Brave's crawler guidance is relevant to the available source pool. It states that Brave's crawler avoids pages that Googlebot cannot crawl. It also says the crawler does not advertise a distinct user agent.

The same guidance distinguishes crawling controls from removal from the index: a site owner can request delisting through a robots noindex directive. Brave must fetch the page again before applying that change. This makes the page's crawl access and indexing directives concrete checks when reviewing its availability through Brave's search infrastructure.

FAQ

Which service can I use with my own model?

Brave's LLM Context documentation recommends that service for applications controlling their own model and processing retrieved page content.

Can I request a completed answer with sources?

Brave's Answers guide describes this workflow. Inline citations use enable_citations=true, and the guide requires streaming for citations and research mode. Research mode allows sequential searches while composing an answer.

Sources

Check how a page reads to answer engines. Open the GEO score analyzer