Story perspectives

AI Chatbots Misrepresent News 45% of the Time

10/22/2025

Summary

Story summary

OpenAI's ChatGPT and Google's Gemini were found by the British Broadcasting Corporation (BBC) and the European Broadcasting Union (EBU) to misrepresent news in 45% of their responses.
The study analyzed over 2,700 AI-generated answers from 22 public media outlets across 18 countries.
Gemini exhibited the highest error rate at 76%.
31% of responses lacked proper attribution.
Experts urge improved media literacy and greater accountability from tech firms to curb misinformation.

History

Historical Context

The 2016 U.S. presidential election saw a surge in fake news, with studies revealing that 44% of U.S. adults relied on Facebook for news. Investigations found that nearly 40% of content from far-right pages was misleading. This misinformation campaign, partly attributed to Russian interference, highlighted the urgent need for media literacy and accountability from tech companies like Facebook and Google to combat ongoing misinformation challenges.

1 of 2

Story summary

OpenAI's ChatGPT and Google's Gemini were found by the British Broadcasting Corporation (BBC) and the European Broadcasting Union (EBU) to misrepresent news in 45% of their responses.
The study analyzed over 2,700 AI-generated answers from 22 public media outlets across 18 countries.
Gemini exhibited the highest error rate at 76%.
31% of responses lacked proper attribution.
Experts urge improved media literacy and greater accountability from tech firms to curb misinformation.

1 / 2