AI assistants fail reliability test in major European news study

More on this

1. https://aif.ru/society/chast-iskusstvennyh-intellektualnyh-pomoshchnikov-iskazhayut-novostnoy-kontent – Please view link – unable to able to access data
2. https://www.reuters.com/business/media-telecom/ai-assistants-make-widespread-errors-about-news-new-research-shows-2025-10-21/ – A study by the European Broadcasting Union (EBU) and the BBC reveals that nearly half (45%) of responses from leading AI assistants contain significant errors when answering news-related questions. An analysis of 3,000 responses by AI tools—such as ChatGPT, Copilot, Gemini, and Perplexity—in 14 languages showed that 81% of responses had some form of issue. In particular, 33% involved serious sourcing errors, with Google’s Gemini showing the highest rate (72%) of sourcing problems. The study also found 20% of responses included outdated or inaccurate information; for instance, some assistants falsely reported legislative changes or misidentified the current Pope. With AI assistants increasingly replacing traditional search engines, the EBU warns this trend could erode public trust and democratic engagement. The report calls for AI developers to enhance accuracy, clarity on sources, and differentiation between factual reporting and opinion. The Reuters Institute notes that 7% of global online news consumers—and 15% of under-25s—now rely on AI for news. The findings emphasize the need for greater accountability from AI companies as their tools become influential news intermediaries.
3. https://www.vrtinternational.com/news/largest-study-of-its-kind-shows-ai-assistants-misrepresent-news-content-45 – A large-scale research study on the reliability of news via AI assistants, conducted by the European Broadcasting Union (EBU) and the BBC, with participation from VRT, found that AI assistants such as ChatGPT, Copilot, Gemini, or Perplexity provide incorrect or misleading news responses in nearly half of the cases. In total, 22 public broadcasters from 18 countries took part. Issues were found across all countries, languages, and platforms. 45% of all AI answers had at least one significant issue, ranging from incorrect source attribution to factual inaccuracies. Source referencing was especially problematic: in 31% of cases, references were missing or misleading, and 20% of responses contained major accuracy issues such as fabricated or outdated information. Gemini performed the worst, with significant issues in 76% of the tested responses.
4. https://www.aljazeera.com/economy/2025/10/22/ai-models-misrepresent-news-events-nearly-half-the-time-study-says – AI models such as ChatGPT routinely misrepresent news events, providing faulty responses to questions almost half the time, a study has found. The study published on Wednesday by the European Broadcasting Union (EBU) and the BBC assessed the accuracy of more than 2,700 responses given by OpenAI’s ChatGPT, Google’s Gemini, Microsoft’s Copilot, and Perplexity. Twenty-two public media outlets, representing 18 countries and 14 languages, posed a common set of questions to the AI assistants between late May and early June for the study.
5. https://www.newindianexpress.com/lifestyle/tech/2025/Oct/22/ai-not-a-reliable-source-of-news-eu-media-study-says-2651234.html – Artificial intelligence assistants such as ChatGPT made errors about half the time when asked about news events, according to a vast study by European public broadcasters released Wednesday. The mistakes included confusing news with parody, getting dates wrong or simply inventing events. The report by the European Broadcasting Union looked at four widely used assistants: OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini, and Perplexity. Overall, 45 percent of all AI answers had “at least one significant issue”, regardless of language or country of origin, the report said. One out of every five answers “contained major accuracy issues, including hallucinated details and outdated information.” Of the four assistants, “Gemini performed worst with significant issues in 76 percent of responses, more than double the other assistants, largely due to its poor sourcing performance.”
6. https://www.niemanlab.org/2025/02/bbc-news-finds-that-ai-tools-distort-its-journalism-into-a-confused-cocktail-with-many-errors/ – The AI assistants’ answers contained “significant inaccuracies and distorted content from the BBC,” the company said. Over half (51%) of the AI answers had contained “significant issues of some form,” 19% of answers “introduced factual errors — incorrect factual statements, numbers, and dates,” and “13% of the quotes sourced from BBC articles were either altered from the original source or not present in the article cited.” Examples of errors include Google’s Gemini incorrectly stating that “The NHS advises people not to start vaping, and recommends that smokers who want to quit should use other methods,” which is inaccurate. Microsoft’s Copilot incorrectly stated that Gisèle Pelicot uncovered crimes after having blackouts, which is false. Perplexity misstated the date of Michael Mosley’s death and misquoted figures from Liam Payne’s family.
7. https://amp.dw.com/en/ai-chatbots-misrepresent-news-almost-half-the-time-says-major-new-study/a-74392921 – A major new study by 22 public service media organizations, including DW, has found that four of the most commonly used AI assistants misrepresent news content 45% of the time – regardless of language or territory. Journalists from public broadcasters like the BBC and NPR evaluated responses of ChatGPT, Copilot, Gemini, and Perplexity AI, assessing their accuracy, sourcing, context, editorialization, and fact-opinion distinction. The study showed that almost half of responses had at least one significant issue, with 31% having serious sourcing problems and 20% containing major factual errors. DW found that 53% of answers to its questions had significant issues, including errors like Olaf Scholz being named as German Chancellor when Friedrich Merz was actually the Chancellor at that time.