The AI news accuracy research represents the largest study of its kind, involving 22 public service media organizations from 18 countries working collaboratively. Professional journalists evaluated AI-generated responses against key criteria including factual accuracy, source attribution, ability to distinguish opinion from fact, and provision of proper context.
The findings raise serious concerns about AI news accuracy as these tools become primary information gateways for millions of users. According to the Reuters Institute Digital News Report 2025, 7% of all online news consumers now use AI assistants for news, with usage rising to 15% among people under 25, indicating a significant generational shift from traditional search engines.
Sourcing errors emerged as one of the most critical AI news accuracy problems identified in the study. Approximately 31% of responses showed serious sourcing issues including missing citations, misleading attributions, or incorrect references to original content. These attribution failures risk undermining trust in both AI systems and the news organizations whose content gets misrepresented.
Major factual inaccuracies represented another significant AI news accuracy concern, with 20% of responses containing serious problems. These included hallucinated details, information generated by AI that never appeared in source materials, outdated information presented as current, and incorrect statements regarding dates, figures, and factual claims.
Google’s Gemini performed worst in the AI news accuracy evaluation, with significant issues appearing in 76% of its responses, more than double the error rate of competing assistants. The primary driver of Gemini’s poor performance was sourcing problems, with 72% of its responses exhibiting citation issues compared to under 25% for ChatGPT, Copilot, and Perplexity.
The AI news accuracy study builds on earlier BBC research from February 2025 that examined the same assistants in English only. That initial investigation found 51% of AI-generated answers to news questions had substantial issues, while 19% of responses referencing BBC material contained factual inaccuracies and 13% of attributed quotes were either modified or entirely fabricated.
Comparing the February findings to the current international study shows some improvements in AI news accuracy, but a considerable number of errors persist across languages and territories. The consistency of problems regardless of geographic location or language suggests systemic issues in how AI assistants process and present news information.
Industry responses to the AI news accuracy findings varied, with companies acknowledging ongoing challenges. OpenAI and Microsoft have previously recognized “hallucinations”, instances where AI generates incorrect or misleading information due to insufficient data, and stated they are working to mitigate these issues. Google indicated it welcomes user feedback to improve Gemini, while Perplexity highlighted a “Deep Research” mode claiming 93.9% factual accuracy.
Public trust in AI news accuracy remains mixed according to companion audience research. A BBC study from October 2025 found just over one-third of UK adults trust AI to produce accurate news summaries, rising to almost half among under-35s. Concerningly, many users assume AI summaries are accurate, and when errors appear, they blame both news providers and AI developers, risking collateral damage to established news brands.
The AI news accuracy study’s authors emphasized the urgent need for transparency in how AI assistants process and present news content. They warned that the growing popularity of these tools could blur lines between verified journalism and synthetic information, potentially undermining public trust in legitimate news organizations that invest in accuracy and accountability.
Investigators called for improved standards and greater accountability among AI developers as these technologies increasingly replace traditional news sources for younger audiences. The study underscores systemic challenges requiring industry-wide solutions rather than incremental improvements from individual companies.
Stay informed about critical research exposing AI limitations and the ongoing efforts to ensure reliable information delivery, visit ainewstoday.org for comprehensive coverage of AI accuracy studies, platform accountability initiatives, and evolving standards protecting journalism integrity in the artificial intelligence era!
