Some 72% of responses by Gemini, Google's AI assistant, had significant sourcing issues, compared to below 25% for all other assistants, it said.

Issues of accuracy were found in 20% of responses from all AI assistants studied