Can Brave LLM Context snippets contain serialized tables or code?

Definition: Yes. Brave's LLM Context API documents snippets that may contain plain text or JSON-serialized structured data, including tables, schemas, and code blocks. The response places these snippets inside content items under grounding.generic.

Key facts

• Brave says LLM Context can extract text, Markdown, JSON, data tables, and code for an agent or model. Its LLM Context documentation names these as use cases.
• The same documentation says snippets may contain plain text or JSON-serialized structured data, including tables, schemas, and code blocks.
• Brave's LLM Context API reference describes grounding.generic as an array of context items with extracted web content.
• The response has separate grounding content and sources metadata, so a snippet is associated with a source URL and title in the response shape.

How are mixed snippets represented?

Brave's LLM Context documentation describes a search endpoint that returns extracted page chunks and source metadata for an LLM. The documented response organizes extracted web content under grounding, with grounding.generic holding URL objects. Each object has a url, a title, and a snippets array.

The example uses quoted strings for the snippet items. A string can carry ordinary prose, while the same field can carry JSON-serialized structured data when Brave extracts a table, schema, or code block. The documented warning therefore matters to post-processing: a consumer needs to accept both plain text and serialized structure.

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

What does the API limit?

Brave's API reference defines maximum_number_of_snippets as the maximum number of different snippets, or chunks of text, included in LLM context. Its default is 50 and its maximum is 256. The same reference defines maximum_number_of_tokens_per_url with a default of 4096 and a maximum of 8192. These controls limit the amount of retrieved context, regardless of whether a snippet contains prose or serialized structure.

The documentation's changelog labels a 2026-07-31 update stating that maximum_number_of_snippets and maximum_number_of_snippets_per_url no longer constrain the response when the token budget allows more snippets. The token budget is therefore the controlling limit in that documented case.

What should a website owner infer?

The LLM Context documentation establishes that Brave's extractor can return tables, schemas, and code blocks as JSON-serialized data inside snippets. It does not publish a page-side markup recipe that makes a particular table or code block appear in a particular snippet. A site review can therefore check whether the source page exposes the information as readable text, a table, a schema, or a code block, while the API consumer prepares for either plain text or serialized structured data.

Related terms

• Brave Search API
• Brave Answers API
• Structured data
• Retrieval-augmented generation (RAG)

FAQ

Can one response contain both plain text and structured snippets?

Yes. Brave's LLM Context documentation explicitly describes a mixed format and says post-processing should be prepared for both plain text and JSON-serialized structured data.

Is a table returned as a native table object?

The same documentation describes the structured content as JSON-serialized data inside a snippet. The response schema separately shows snippets as an array of extracted text chunks, so the documented consumer contract is an array of snippet values rather than a separate table field.

Can the API return code as part of a snippet?

Yes. The same documentation lists extracting code to answer technical questions as a LLM Context use case and includes code blocks among the structured data that snippets may contain.

Sources

1. Brave, "LLM Context: Web search for AI agents and chatbots", undated
2. Brave, "LLM Context: Web search API for AI agents", undated

---

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