How AI Chatbots Spread Misinformation During Online Searches

Tech / Tech / AI
By Newsroom,  published 24 October 2025 at 19h37, updated on 24 October 2025 at 19h37.
Tech

ADN

As artificial intelligence chatbots and digital assistants become increasingly integrated into our online searches, concerns are rising about the spread of misinformation. These tools, while innovative, sometimes deliver inaccurate or misleading information to users.

TL;DR

  • AI assistants often provide inaccurate or misleading information.
  • Most struggle with sourcing and factual accuracy issues.
  • Users must verify AI-generated news with primary sources.

AI Assistants in Daily Life: Ubiquitous but Flawed

Artificial intelligence has quietly slipped into our routines, shaping the way we search for information, keep up with current events, and, increasingly, how we interpret the news. Whether it’s ChatGPT, Google Gemini, or other platforms, these tools are ever-present. Yet, despite their growing popularity, a recent study by the European Broadcasting Union (EBU) raises troubling questions about their reliability.

The EBU Study: Widespread Errors and Inconsistent Sourcing

Researchers at the EBU examined over 3,000 queries across a range of popular AI assistants, including Microsoft Copilot, Claude, and Perplexity. The scope was broad—testing performance in no fewer than 14 languages. Their findings paint a sobering picture:

  • 31% of responses contained sourcing issues such as non-existent or incorrect references.
  • 20% of answers displayed significant factual errors—including misdated events or misattributed quotes.

Some platforms, like Gemini, struggled to accurately cite their sources; others showed variable quality depending on the version in use. Strikingly, nearly 45% of all replies included at least one major mistake, and a staggering 81% exhibited some kind of problem—from outdated data to misleading language.

Misinformation on Demand: A Challenge for Audiences and Media Alike

According to research from the Reuters Institute, roughly 15% of Generation Z already turns to chatbots for news updates. This shift brings risks: many assistants blend fact with opinion or fail to disclose their sources clearly. Take a simple test—ask multiple AIs about the “latest U.S. debt ceiling agreement.” Responses vary wildly. Only Claude delivered an answer that was both clear and accurate; others—such as ChatGPT—sometimes cited fictitious articles dated in the future, undermining trust.

How can users protect themselves from these pitfalls? Several factors explain this need for caution:
– Frame questions specifically (for example: “Provide recent, reliable source links”).
– Request precise timestamps (“As of October 23rd, what is the status?”).
– Always cross-check answers among several AI models.
– Never rely solely on summaries without reviewing primary sources.

Ultimately, while AI assistants offer convenience, they do not replace diligent human verification. As audiences increasingly migrate toward automated interfaces—often at the expense of traditional publishers—the collective trust in information is at stake. Until real-time transparency and accuracy become standard, returning to original reporting remains a wise choice.

Learn more