Brave Search Help

AI in Brave Search

Brave Search utilises state-of-the-art artificial intelligence to enhance your search experience and provide you with the most useful results. Advanced machine learning models are used broadly and deeply across most parts of the technology stack that powers Brave Search. Large language models (LLMs) are also used to synthesize information into more digestible bits. Some of our user-facing AI-powered features include:

  • Answer with AI: When you ask Brave Search a question, we provide a concise summary answer along with references to sources to back up the key claims. Provenance and transparency are crucial, so we always show you where the information comes from. Parts of this feature are built with open source LLMs such as Meta Llama 3, Mistral (and Mixtral). Read the announcement.
  • CodeLLM: CodeLLM is aimed at queries that involve programming, and combines the depth and quality of search results with the summarization and explainability power of LLMs. CodeLLM offers programmers an option to get AI code snippets grounded in search results, with step-by-step explanations and citations of the sources for reference and validation. Read the announcement.
  • Featured Snippets: For certain questions, one source may do an excellent job of directly answering your query. In these cases, we analyze the page and extract the most relevant snippet of text that answers your question. This helps you get the information you need as quickly as possible.
  • AI-powered descriptions: Where possible, we generate descriptions for search results using a Question Answering model. This provides you with a helpful brief on what the results contain and their relevance to your search, saving you time clicking and scanning.

Brave Search AI is built in-house and powered by our own models and algorithms. Our goal is to provide you with independent, private search experiences that put you in control.