In AI chatbots there is a lack of rigor when it comes to reporting current news

Ubiquitous AI chatbots are revolutionizing the way we get information on the Internet. However, In the responses provided by this type of platforms, a notable lack of rigor is often seen, particularly when the queries are directly related to current news.. This is what at least a recent report from the European Broadcasting Union (EBU) concludes.

To carry out the investigation The answers provided by ChatGPT, Copilot and Perplexity to questions about current news were put under the microscope. In total, more than 3,000 responses were examined and graded based on accuracy, references to sources used, differentiation between factual information and opinion, and provision of appropriate context to the content.

The results of the study undertaken by the EBU are inevitably alarming (especially if we take into account the growing number of users who throw themselves into the arms of ChatGPT and company when it comes to stocking up on current information) and they definitely do not leave AI chatbots in a good place.

Accuracy is the Achilles heel of AI chatbots when it comes to reporting current news

Almost half of the responses that come from the lips of AI chatbots (45%) contain at least one major error. And the most common errors are directly related to the sources cited by AI chatbots. 31% of the responses show, not in vain, serious problems when referring to the sources of information on which they anchor. Gemini, Google’s AI chatbot, scores particularly poorly on this point with a 72% error rate when citing sources. Similarly, one-fifth of AI chatbots (20%) fail to provide the user with truly accurate responses, while 14% bypass necessary context in responses.

Anyway, AI platforms appear to have made progress over the past few months in responding to queries about current news. The EBU report is based on an investigation published by the BBC in February 2025. And that first report showed even more worrying results.

While in the first study half of all the answers examined (51%) contained significant errors of different types, in the second investigation this proportion dropped to 37%.

In light of the discouraging results put on the table by the EBU report, The European Broadcasting Union and its members are calling on national and European authorities to apply the rules currently in force in relation to the diversity of digital services and media. According to the EBU members, independent monitoring of AI tools is necessary in view of the absolutely rapid development of this technology.

The EBU has also published a toolkit with suggestions to ensure the highest possible accuracy in the responses provided by AI chatbots.

They took part in the EBU investigation 22 public broadcasting corporations from 18 different countries (including RTVE).