ChatGPT search, OpenAI’s feature within ChatGPT that allows chatbots to access relevant information from the web and incorporate it into their responses, is growing rapidly in Europe.
A report filed by one of OpenAI’s EU corporate entities, OpenAI Ireland Limited, shows that ChatGPT searches had approximately 41.3 million average monthly active “recipients” in the six-month period ending March 31. This compares to approximately 11.2 million average monthly active recipients for the six months ended October 31, 2024.
OpenAI regularly publishes search information in ChatGPT to comply with the EU’s Digital Services Act (DSA), which regulates many aspects of online services in European countries. The DSA defines monthly active recipients as “people who actually interact with a service at least once in a given period of time” by “being exposed to information disseminated in the interface of the online platform, for example by viewing it or listening to it, or by providing information”.
One component of the DSA mandates that “very large” online platforms or search engines-those with more than 45 million average monthly recipients-allow users to opt out of recommendation and profiling systems, share certain data with researchers and authorities, and be subject to external audits. ChatGPT search may soon meet these requirements if the current growth trend continues.
Online platforms that fail to comply with DSA rules can be fined up to 6% of their global turnover. A platform that persistently refuses to comply with the rules may face a temporary suspension from operating in the EU.
Since its debut last year, ChatGPT Search has begun to make inroads on leaders such as Google. According to a survey published in September, 8% of people said they would choose ChatGPT over Google as their primary search engine. But Google remains the dominant online search tool. According to one estimate, it processes 373 times more queries than ChatGPT.
Researchers have found that searches on ChatGPT and other AI-powered search engines are less reliable than regular searches, depending on the query. According to one study, ChatGPT misidentified 67% of the articles it searched for. Another study found accuracy issues related to ChatGPT’s handling of news content, including content from publishers with whom OpenAI has licensing agreements.
