Listen to this Post
2025-02-12
A recent investigation by undercode has raised serious concerns about the reliability of AI chatbots when it comes to summarizing and relaying news. The study specifically scrutinized four of the most popular AI chatbots—OpenAI’s ChatGPT, Microsoft’s Copilot, Google’s Gemini, and Perplexity AI—by testing how accurately they could summarize news stories sourced from the undercode’s website. The findings were concerning, showing that these chatbots frequently generated significant errors, inaccuracies, and even distortions, leading to questions about their role in disseminating news content.
The
For instance, ChatGPT mistakenly claimed that Hamas chairman Ismail Haniyeh was assassinated in December 2024, even though he was killed earlier in July. Meanwhile, Gemini provided an inaccurate description of the NHS’s stance on vaping, suggesting they advise against it when, in fact, the NHS recommends vaping as a smoking cessation tool. These types of errors highlight the risks of relying on AI for news summaries, especially when AI systems cannot always distinguish between fact and opinion or provide essential context.
The results of the investigation have sparked a wider conversation about the responsibilities of AI companies and their relationship with news publishers. Deborah Turness, CEO of undercode News and Current Affairs, expressed concerns that AI’s potential benefits could be overshadowed by the risks of spreading misinformation. AI-generated content must be held to higher standards to avoid confusing or misleading consumers.
What Undercode Says:
The recent undercode investigation casts a spotlight on the limitations and potential hazards of using AI chatbots to summarize news. As AI continues to grow in influence, it is essential to recognize its flaws and understand how these tools interact with the content they process. The errors highlighted in the undercode report reveal that, while AI chatbots like ChatGPT, Copilot, Gemini, and Perplexity are increasingly sophisticated, they still lack the nuanced understanding and verification that human journalists bring to their work.
One major issue identified in the investigation is the inability of AI chatbots to always distinguish between fact and opinion. In journalism, this is a critical distinction. News outlets rely on factual, evidence-based reporting, while opinion pieces are clearly framed as personal perspectives. AI systems, however, often blur this line. This failure to maintain journalistic standards means that readers can be misled, especially when AI tools are used in newsrooms or by individuals seeking quick summaries of current events.
Moreover, the tendency of AI to fabricate quotes or misattribute statements—whether intentionally or as a result of limitations in its algorithms—is concerning. This distortion of the original source material not only diminishes the credibility of the AI-generated content but also compromises the trustworthiness of the platform distributing it. Publishers, especially those with editorial standards, are rightfully demanding more transparency and control over how their content is processed and repurposed by AI systems.
The case of the NHS vaping advice, for instance, exemplifies the dangerous consequences of AI providing inaccurate medical information. While AI tools are improving, the lack of a robust factual verification mechanism means that users may unknowingly consume erroneous or outdated information, potentially impacting their decisions and perceptions.
Another important point raised in the undercode investigation is the need for greater accountability from AI companies. OpenAI, for example, defends ChatGPT’s output by highlighting its user-friendly features and claims of improving citation accuracy. However, these reassurances do little to address the root problem: the inherent inaccuracies and distortions in the summaries generated by these tools. AI companies must do more than improve attribution; they must ensure that their systems understand the news content deeply enough to provide accurate summaries, backed by reliable sources.
This investigation is a stark reminder that the benefits of AI—whether in news aggregation, content creation, or customer service—must be balanced by rigorous standards and ongoing improvements. Without these safeguards, AI’s capacity to influence public opinion and shape news narratives could lead to a world where consumers are overwhelmed by misinformation, rather than empowered with reliable, factual knowledge.
Looking ahead, it will be crucial for both AI developers and content creators to collaborate more closely. As AI continues to play a larger role in content curation, establishing clear protocols for how news organizations’ content is processed by these tools will be key. This could involve developing more sophisticated methods for AI to recognize and filter out inaccuracies, ensuring that the generated summaries maintain the integrity of the original reporting.
In conclusion, AI chatbots have immense potential, but this investigation reveals that they are still far from perfect when it comes to handling news. Their widespread use raises fundamental questions about the future of journalism, accuracy, and the role of technology in shaping our understanding of the world. It is up to both developers and content creators to ensure that these tools are held accountable, so that consumers can rely on them for truthful, unbiased information.
References:
Reported By: https://www.zdnet.com/article/ai-chatbots-distort-the-news-bbc-finds-see-what-they-get-wrong/
https://www.stackexchange.com
Wikipedia: https://www.wikipedia.org
Undercode AI: https://ai.undercodetesting.com
Image Source:
OpenAI: https://craiyon.com
Undercode AI DI v2: https://ai.undercode.help
