OpenAI tests AI search engine SearchGPT

The AI company OpenAI is currently testing a search function based on artificial intelligence (AI). SearchGPT is integrated directly into ChatGPT. However, the function can initially only be tried out by testers, including individual publishers. It is basically an extension of the web search in the chatbot. Because it is a search function, the new crawlers from OpenAI access all websites. The exclusions of other bots therefore do not apply here.

However, the AI search function does not yet appear to be fully developed, as a video shown in the blog post reveals an error in the answers to the search query for music festivals taking place in the city of Boone in the US state of North Carolina in August. SearchGPT returns a list of events with a short description and a link to the source. Even the first entry is incorrect and gives an incorrect event date. The AI search therefore still has its pitfalls.

Nevertheless, links are displayed much more prominent than was the case with ChatGPT's previous answers. This brought and still brings OpenAI a lot of criticism that it was not clear where information came from.

SearchGPT is not yet available to all users anyway. However, there is an option to be placed on a waiting list. In addition to improvements to the search function, OpenAI is also planning to incorporate more local information and commerce. This sounds suspiciously like Google's classic search, which prominently displays a lot of sponsored content and eCommerce. It is not yet known whether OpenAI also plans to display advertising. Other search engine providers finance their services this way. Paid search engines have mostly failed. AI applications are generally expensive to develop and operate – OpenAI's business model is already on shaky ground.

More relevant search results

OpenAI believes that SearchGPT can simplify the search for answers on the internet and deliver more relevant results. "We believe that by extending the conversational capabilities of our models with real-time information from the Internet, finding what you're looking for can become faster and easier," says OpenAI. It was also previously possible to have ChatGPT search the web in real time. The results were just displayed differently.

OpenAI had discontinued the original web search in the meantime because of a dispute with the New York Times. The publisher has since filed a lawsuit for copyright infringement. ChatGPT is said to have displayed articles that were behind paywalls. Such cases should no longer occur. In the case of AI models, there is the question of copyright and the use of data for training the models. On the other hand, there is also the issue of ancillary copyright, i.e. the question of how information is used and who the creators are referred to. Google and Meta have been battling with regulatory authorities worldwide for years.

OpenAI claims to have worked closely with publishers and authors – also to avoid legal problems. According to OpenAI, publishers are provided with a tool that allows them to manage how their information appears on SearchGPT. This applies to all publishers. However, the linked support page initially only leads to an overview of the crawlers that OpenAI uses. The relevant bot is described there: OAI-SearchBot is used exclusively for searching and not for training AI models.

With SearchGPT, OpenAI is preparing to enter into direct competition with Google's search engine, which dominates the search engine market with over 90 percent. Whether this will work is uncertain. Previous attempts by smaller AI search engines to take a share of Google's search engine business have failed.

(olb)