OpenAI evaluates SearchGPT for transforming web searches into valuable answers

OpenAI has begun testing its new search engine, SearchGPT, aiming to avoid issues faced by competitors. This "temporary prototype" provides timely web information, allows for follow-up questions, and prominently cites its sources.

OpenAI is working with creators and publishers to enhance website and experience discovery, with The Atlantic as one example. They are also seeking user feedback on the prototype.

Importantly, SearchGPT will not use data to train OpenAI's generative models like ChatGPT or DALL-E, and websites that opt out of training won't be penalized.

To test SearchGPT, users must join a waiting list. OpenAI aims to integrate this improved search experience into ChatGPT and enhance local and business searches.

This strategy is designed to address criticisms faced by other AI-driven search engines, like Perplexity and Google's AI Overviews, which have been accused of content scraping and not directing users to original sources. While SearchGPT may not fully resolve these issues, OpenAI’s approach reflects an effort to acknowledge and address them, especially in light of potential regulatory changes regarding AI and copyright.