SAN FRANCISCO, Sept. 19, 2024 (GLOBE NEWSWIRE) — Andi, which is building search for the next generation, announced today its top-ranking performance in the inaugural SearchBench AI benchmark, a new independent evaluation of AI search accuracy developed by Talc. In a field including industry giants like Google Gemini and OpenAI’s ChatGPT, Andi achieved the highest score for factual accuracy, demonstrating the power of its innovative approach to AI search.
SearchBench AI, launched today by Y Combinator-backed Talc, represents a significant leap forward in evaluating AI search engines. Unlike traditional static benchmarks, SearchBench AI generates 900 challenging questions based on current events and live web data. This state-of-the-art approach prevents AI companies from gaming the system by training on known datasets. This SearchBench AI evaluation ran from August 30, 2024 to Sept 2, 2024. Andi shared common issues for GenAI accuracy with the Talc team – this helped Talc identify key areas to focus on.
This is the first comprehensive real world benchmarking of accuracy in AI Search.
“SearchBench AI is designed to assess how AI search performs in real-world scenarios,” said Matt Lee, co-founder and CEO of Talc. “We wanted to create a practical benchmark that focuses on the everyday usefulness of these tools, not just academic capabilities.”
The SearchBench AI results showed:
- Andi – 87% accuracy
- You.com – 80% accuracy
- Google Gemini – 71% accuracy
- OpenAI ChatGPT – 62% accuracy
- Perplexity AI – 59% accuracy
Full benchmark data is available at https://github.com/Talc-AI/search-bench.
“These results validate our unwavering focus on factual accuracy,” said Angela Hoover, co-founder and CEO of Andi. “We’ve built Andi from the ground up to understand content deeply, not just match keywords. This approach allows us to provide users with trustworthy, spam-free results – a critical need in today’s information landscape.”
Andi’s performance is particularly notable given the company’s size and resources compared to its competitors. With just $3 million in funding, Andi has developed proprietary technology that outperforms solutions from companies with billions in resources.
“Our team has been laser-focused on delivering accurate results to real user queries,” added Hoover. “Every innovation we make is driven by the goal of helping people find reliable information quickly and easily.”
The Rise of Factual AI Search
As AI-powered search becomes more prevalent, concerns about accuracy and misinformation have grown. Traditional search engines, driven by ad revenue, often prioritize engagement over accuracy. Meanwhile, many AI models struggle with “hallucinations” – confidently stating false information.
Andi addresses these issues through its proprietary Trantora backend, which analyzes the meaning and credibility of web content, not just keywords. This approach allows Andi to provide concise, accurate answers free from ads, spam, and misinformation.
“In the age of AI, trust is everything,” said Hoover. “Users need to know they’re getting reliable information. Our performance in the SearchBench AI benchmark shows that with the right approach, AI can dramatically improve everyday search accuracy, not compromise it.”
Industry Impact
Andi’s strong performance in an independent benchmark has caught the attention of industry experts.
“What Andi has achieved is remarkable,” said Lorenzo Thione, General Partner at Gaingels, an investor in Andi. “These results show that innovative startups can solve problems that have plagued even the largest and most funded tech companies. This could be a turning point for trust and adoption in AI search.”
Looking ahead, Andi plans to continue refining its technology and expanding its capabilities, always with a focus on delivering accurate, trustworthy results to users.
“We’re just getting started,” said Hoover. “Our mission is to build a search engine that truly understands and serves the needs of users. This benchmark is a milestone, but there’s much more work to do to get to 100% accuracy and trust.”
For more information about Andi and to experience its accurate AI search firsthand, visit andisearch.com.
About Andi
Andi is search for the next generation: visual, conversational and factually accurate. Instead of just links, Andi gives you answers – like chatting with a smart friend. Search is broken because of spam and ad tech. Andi uses generative AI and large language models combined with live data, smart algorithms and semantic search technology – with no spam or ads. Named a top 3 AI search engine by Fast Company, Andi is focused on providing factually accurate and concise answers with a visual feed of the most relevant, high-quality sources. More than 1M people already use Andi – try it for yourself at andisearch.com.
Media Contact:
Michelle Faulkner
Big Swing
michelle@big-swing.com
617-510-6998
A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/d6f56fe1-103b-455a-b490-b4cbb4f1f8bc
GenAI search engine Andi bested Google, OpenAI and others in a benchmark study on factual accuracy
Named a top 3 AI search engine by Fast Company, Andi is focused on providing factually accurate and concise answers with a visual feed of the most relevant, high-quality sources. More than 1M people already use Andi.
COMTEX_457915876/2471/2024-09-19T09:00:08