![Data Confirms Disruptive Potential Of SearchGPT Data Confirms Disruptive Potential Of SearchGPT](https://www.searchenginejournal.com/wp-content/uploads/2024/08/searchgpt-781.jpg)
Researchers analyzed SearchGPT’s responses to queries and identified how it may impact publishers, B2B websites, and e-commerce, discovering key differences between SearchGPT, AI Overviews, and Perplexity.
What is SearchGPT?
SearchGPT is a prototype natural language search engine created by OpenAI that combines a generative AI model with the most current web data to provide contextually relevant answers in a natural language interface that includes citations to relevant online sources.
OpenAI has not offered detailed information about how SearchGPT accesses web information. But the fact that it uses generative AI models means that it likely uses Retrieval Augmented Generation (RAG), a technology that connects an AI language model to indexed web data to give it access to information that it wasn’t trained on. This enables AI search to provide contextually relevant answers that are up to date and grounded with authoritative and trustworthy web sources.
How BrightEdge Analyzed SearchGPT
BrightEdge used a pair of search marketing research tools developed for enterprise users to help identify search and content opportunities, emerging trends and conduct deep competitor analysis.
They used their proprietary DataCube X and the BrightEdge Generative Parser™ to extract data points from SearchGPT, AI Overviews and Perplexity.
Here’s how it was done:
“BrightEdge compared SearchGPT, Google’s AI Overviews, and Perplexity.
To evaluate SearchGPT against Google’s AI Overviews and Perplexity, BrightEdge utilized DataCube X alongside BrightEdge Generative Parser ™ to identify a high-volume term and question based on exact match volumes. These queries were then input into all three engines to evaluate their approach, intent interpretation, and answer-sourcing methods.
This comparative study employs real, popular searches within each sector to accurately reflect the performance of these engines for typical users.”