With a new ‘answer engine,’ Brave browser adds another generative AI tool for search
By Marty Swant • April 17, 2024 • 4 min read •
Ivy Liu
Brave, a privacy-focused browser, is rolling out a new generative AI tool to give users enter a single query to get real-time information from multiple sources.
“AI Answer Engine,” which debuted today, provides a synthesized summary with source citations alongside traditional search results. But instead of using Google or Bing, the tool uses Brave’s independent search index, which the startup claims is “curated and cleansed of SEO spam and junk content.”
Platforms like Google and Bing are also expanding capabilities for AI search, but Brave aims to put its privacy-preserving features at the forefront. Because the company doesn’t collect user queries or create session profiles for ad-retargeting, Brave Chief of Search Josep Pujol said AI answers are based on the results of a user’s query rather than their user profile or search history.
That doesn’t mean Brave won’t consider adding ads at some point. Pujol said Answer With AI is meant to complement traditional search, not replace it. Brave already has its own advertising business, which includes ways for users to make their own revenue from ads through its Brave Rewards program.
“Integrating answers on the normal search engine result page has proven useful over many years,” Pujol told Digiday via email. “Thus, ads could potentially be featured in the area that is taken by the answers. It is not the case at the moment but it could, same as for the results area. Ads are always labeled as such so that the user clearly sees what is sponsored and what is organic, and it would be the same case for Answer with AI.”
When developing its AI search tool, Brave focused on popular ways people search based on types of intent. Common categories include navigating to specific websites, looking for information, researching products and services and “transactional” actions like buying a product. According to Brave, the search index includes more than 1 billion location-based schemas that help pull information from more than 100 million businesses.
Brave’s approach uses retrieval augmente generation (RAG), an increasingly popular technique for improving the accuracy and relevancy of responses from AI chatbots.
Results from AI Search Engine are grounded by answers on pages indexed by Brave Search, which has been using RAG for a year through an earlier tool called Summarizer. According to Pujol, Summarizer has already served “hundreds of millions of queries” since its debut in March 2023. He added that Brave is now expanding both the quality and quantity of the data feed as part of the context the model uses to answer each prompt.
Searching beyond the giants
In some ways, Brave’s updates look similar to what the search startup Neeva was developing to compete with Google until it was acquired by Snowflake last May. Since then, Neeva’s LLMs have been part of Snowflake’s strategy for enterprise customers and power a new family of AI models that debuted this week.
Since it debuted its generative AI feature a year ago, Brave also has seen its user base increase from 57.8 million monthly active users in March 2023 to 73.3 million last month. Daily active users have also increased from 21.7 million active users to 26.3 million during the same period.
Brave’s updates arrive as other search startups like The Browser Company and Perplexity also look to gain traction with generative AI. According to Similarweb, visits to Perplexity’s website totaled 61.5 million in March, up from 49.5 million in February and 45.6 million in February.
Generative AI’s effect on search results for users and advertisers is yet to be determined — and likely won’t be for a while. However, advertising’s influence on search has been an age-old debate since the beginning of search itself. Even Google’s cofounders’ original paper questioned whether ads would pollute results. Meanwhile, others worry that more AI-generated misinformation in search results has also been an ongoing concern. Search is also a key factor in the ongoing antitrust case between Google and the U.S. Justice Department.
Startups like Perplexity are also exploring ways to add ads to their generative search results. In a recent emailed statement to Digiday, chief business officer company Dmitry Shevelenko said it wanted to show ads without skewing results. For example, Perplexity doesn’t plan to show ads before answers or let advertisers pay for better placement.
“We will adhere to both of these criteria in developing our ad products, and are exploring using our ‘Related Questions’ feature as a natural entry point for brands to influence the follow-up questions people ask on Perplexity,” according to Shevelenko. “However, this is all speculative at this stage as we have no firm products built yet or timeline on releasing them. We will keep you posted as we finalize our plans and product scope.”
https://digiday.com/?p=541513
