Attention AI users! The results of this research will surprise you...

In an international study conducted jointly by the European Broadcasting Union (EBU) and the UK-based BBC, approximately 3,000 responses were examined by some artificial intelligence assistants that answered users' questions.

Responses were assessed for accuracy, source attribution, and distinction between commentary and information. Twenty-two public broadcasters from 18 countries participated in the survey, and the analysis was conducted in 14 different languages.

In this context, it was determined that approximately 45 percent of artificial intelligence applications such as ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity contained serious errors in their responses, and 81 percent had at least one problem.

According to the research, one-third of the responses contained serious sourcing errors, such as inaccurate, incomplete, or misleading sources. 72% of Google Gemini responses had serious sourcing issues, while the figure for other platforms was below 25%.

In terms of accuracy, 20% of all assistant responses contained incorrect or outdated information. For example, Gemini incorrectly stated that there had been a change to the law on disposable e-cigarettes, while ChatGPT reported that Pope Francis was still in office months after his death.

EBU Media Director Jean Philip De Tender said in a statement, "When people don't know what to trust, they trust nothing, and this can hinder democratic participation." Source: AA This content was published by Enes Çırtlık .