ChatGPT And Gemini Making Errors With AI News Summaries That Should Worry You – News18

Last Updated:February 24, 2025, 11:38 IST

AI news summary was projected as useful feature for millions but companies are now realising these tools have a major problem.

ChatGPT, Google’s Gemini and other popular AI chatbots are growing in popularity but these chatbots have a major weakness and that is summarising news stories. Yes, it is not just Apple but other AI tools are also having trouble in summarising the news for users, according to a new study that came out earlier this month.

AI models are being trained on existing data and we have seen news media organisations partner with the likes of OpenAI for training these AI chatbots. But even then, the AI summaries are not accurate and in some worrying cases, distorted as well.

The AI News Summary Trouble

The findings have come via a study by the BBC where the likes of ChatGPT, Copilot and Gemini were asked to summarise around 100 news stories. The entity got journalists for relevant fields to check these summaries and rate them.

The results were shocking to say the least. The study saw over 50 percent of the AI summaries have some major issues in the content. That’s not all, 19 percent of these AI news stories had factual errors, something that we’ve heard about Apple’s AI summary tool which was eventually disabled with the iOS 18.3 update.

The study clarifies one thing about AI tools that we have heard experts say for a while now: slow down the AI roll out and even pull back features if they make errors. Google has also faced its share of AI lapses and its image generation tool was accused of tweaking features in images with racial bias in play.

Guard rails around AI and the technology is critical and studies like these show that the features are exciting but the level of their accuracy is a major red flag and giants like Google or OpenAI need to be careful how they package these tools before it is made available to the public.

Location :

United Kingdom (UK)