According to The Information, Openai, which Microsoft invests, uses Google search results, although it has no direct access to them. To do this, the Serpapi service is used-a platform that collects search results using a web scraping.
These data help ChatGPT answer questions about current events, for example, about sports results or fluctuations on the exchange. A year ago, Openai directly requested access to Google’s search, but received a refusal. Since then, Google has been trying to complicate the work of Serpapi, although he does not sue, so as not to aggravate his antitrust trial in the United States.
In addition to Serpapi, Openai uses its own web scanner, as well as data from Microsoft Bing and other companies. However, according to the recognition of Openai itself, these sources are not yet enough to achieve the accuracy of Google, especially in case of rare requests.
The developers also noticed that sometimes ChatGPT answers look like ready -made fragments from Google search results. In one of the experiments, a former Google engineer created fake pages that were indexed only in Google. Despite this, ChatGPT was able to display them in answers, which indicates the use of information from there.
