OpenAI has introduced SearchGPT, an AI-powered search engine designed to provide users with fast and timely answers while prominently attributing and linking to publishers in search results. The search engine, powered by the GPT-4 family of models, aims to organize and summarize search results, offering clear, in-line attributions and named links to enable users to identify the sources of information and access additional results with ease through a sidebar featuring source links.

The decision to release SearchGPT as a prototype offers several advantages for OpenAI. It provides a buffer in case of significantly inaccurate results and mitigates the risk of incorrect attributions or potential issues related to content usage. The development of SearchGPT has been a topic of speculation for several months, with OpenAI's strategic focus on building a competitive search platform evident through efforts to recruit Google employees for a search team and the development of a new website hinting at the impending move into the search market.

OpenAI has gradually enhanced ChatGPT's integration with real-time web content, with plans to integrate SearchGPT's features directly into ChatGPT. However, the company faces increasing costs associated with AI training and inference, with projected costs reaching $7 billion this year. SearchGPT will be initially offered for free, and the absence of ads indicates that OpenAI will need to devise a monetization strategy for the feature in the near future.

The launch of SearchGPT marks OpenAI's entry into a competitive search engine market, positioning the company to challenge Google's dominance. The search engine's prototype status allows OpenAI to gather feedback and refine the service, aligning it with the real-time web and providing a potential alternative to existing search engines.