How does Brave LLM Context differ from ordinary web results?

Definition: Brave LLM Context is an API that returns extracted page chunks and source metadata for an AI agent or application. Ordinary Brave Web Search returns result data intended for a human results page.

Key facts

• Brave describes LLM Context as advanced Web search with pre-extracted content for AI agents, LLM grounding, and RAG pipelines in its LLM Context documentation.
• The API response separates extracted content in grounding from metadata in sources, and its API reference describes the latter as keyed by URL.
• Brave's Web Search documentation describes ordinary results as intended for human consumption. Its extra_snippets option can add up to 5 additional excerpts per result.

What does LLM Context return?

The LLM Context guide says its data-first ranking compiles relevant smart chunks into a compact format for LLM consumption. It describes the contrast with traditional Web Search directly: ordinary results provide titles and short snippets, while LLM Context provides extracted page chunks and source metadata. The endpoint is therefore a retrieval response that an application can pass to its own processing step.

The response has two related layers. grounding contains the content selected for the query. In the generic case, each item has a source URL, a page title, and snippets. The API reference defines those snippets as extracted text chunks selected for their relevance to the query.

{
  "grounding": {
    "generic": [
      {
        "url": "https://example.com/page",
        "title": "Page Title",
        "snippets": [
          "Relevant text chunk extracted from the page..."
        ]
      }
    ]
  },
  "sources": {
    "https://example.com/page": {
      "title": "Page Title",
      "hostname": "example.com"
    }
  }
}

What are the source metadata fields?

The API reference describes sources as metadata for all referenced URLs, keyed by URL. A source entry can expose the page title, hostname, and age. The LLM Context guide documents four age positions: a full date, an ISO date, a relative age, and an ISO 8601 timestamp. When no page date is known, the age array is empty.

The optional enable_source_metadata parameter enriches each source entry with site_name, favicon, thumbnail, description, and fetched_content_timestamp. The reference gives false as its default. The description is the page's own description, independent of the query, and the fetched content timestamp records the index's last fetch in Unix seconds.

Snippets can contain plain text or JSON-serialized structured data such as tables, schemas, and code blocks. An application that post-processes the response therefore needs to preserve the snippet value as mixed content rather than assuming that every chunk is prose.

How does this compare with Web Search?

Brave's Web Search page describes a large index with results intended for human consumption. It exposes ordinary result previews and an extra_snippets option that provides up to 5 additional excerpts per result. These excerpts add context to a result card while keeping the response organized as search results.

LLM Context is documented for cases where an agent or model is the recipient. Its response puts query-relevant extracted chunks in grounding.generic, then lets the application join each URL to the corresponding entry in sources. The two APIs can therefore expose related retrieval inputs while presenting different response structures and different intended consumers.

How can you inspect the difference?

For an LLM Context response, inspect grounding.generic[].snippets first, then match each item's url to the same key in sources. Record whether optional metadata is present and whether a snippet is plain text or serialized structured data. The API reference names url, title, snippets, hostname, and age as response fields.

Related terms

• Brave Search API
• Ask Brave
• Brave Answer with AI

FAQ

Does LLM Context generate the answer?

Brave documents LLM Context as raw extracted content for an application's own LLM pipeline. The same documentation describes its separate Answers service as the option for direct AI answers.

Are its snippets the same as ordinary search snippets?

They serve different response shapes. Web Search can add up to 5 extra excerpts per result, while the LLM Context API defines snippets as extracted chunks selected for query relevance.

Is source metadata always enriched?

No. The API reference sets enable_source_metadata to false by default. The optional fields such as description and fetched_content_timestamp require that parameter.

Sources

1. Brave Search API, "LLM Context: Web search for AI agents and chatbots", 2026-07-31
2. Brave Search API, "Web Search", undated
3. Brave Search API, "LLM Context API reference", undated

---

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