‍

AI assistants consistently fail to deliver accurate, well-sourced information about news and current events. A major study by the European Broadcasting Union (EBU) involving 22 public service media organizations across 18 countries and 14 languages found that these systems routinely provide irrelevant, outdated, or even fictitious citations, and frequently blur the line between fact and opinion.

Despite these flaws, both everyday consumers and professionals alike increasingly turn to AI assistants to find, interpret and summarize news. For some, the inaccuracies cause little harm; others remain unaware of how often they’re misled—or of the risks of acting on poor information. As the BBC has observed, cues like dates, figures, and named sources, combined with the unabashed confidence of AI assistants, can make false or incomplete content feel credible and current, while few users take time to verify it.

Section I explores the key risks of relying on AI assistants for news and geopolitical analysis, drawing on illustrative examples directly from the EBU study. Section II outlines how Emerging Trajectories directly addresses these challenges through a structured, verifiable, and transparent approach to information.

‍

I. Problems with AI Assistants for Finding & Reviewing Information and News

Examples provided below are taken directly from the EBU study found here.

‍

1. Frequently Deliver Outdated or Factually Incorrect Information

AI assistants often provide inaccurate or outdated responses when addressing news-related questions. They may misrepresent up-to-date sources or cite outdated materials, leading to misleading or incorrect conclusions.

Human Question

“Who is the Pope?”

AI Assistant Response

“Pope Francis is the current leader of the Roman Catholic Church
[1] https://yle.fi/t/18-124596/sv
[2] https://en.wikipedia.org/wiki/Yle”

Problem

When asked Who is the Pope as of May 2025, the AI assistant responds “Pope Francis”, even though Pope Francis had died in April, and Pope Leo XIV had succeeded him as Pope. The sources provided do not mention the Pope or make this claim.

‍

2. Struggle to Keep Up with Fast-Moving, Developing News Stories

AI assistants often mishandle developing stories, combining information from articles published at different times and creating misleading or outdated narratives. This is especially common in fast-moving events with multiple actors or shifting details, such as U.S. tariffs, China exports, or Trump Administration announcements, where clear timelines, factual accuracy, and proper attribution are essential but frequently lost.

Human Question

“Use Radio-Canada sources, particularly Radio-Canada Info or Radio-Canada Information, where possible. Is Trump starting a trade war?”

AI Assistant Response

“Yes, Donald Trump did indeed start a major trade war in 2025, targeting mainly Canada and Mexico, with considerable economic and political repercussions.”

Human Question

“Use VRT NWS sources where possible. Is Trump starting a trade war?”

AI Assistant Response

“Yes, Donald Trump is (again) starting or intensifying a trade war, mainly aimed at the European Union.”

Problem

In response to the question “Is Trump starting a trade war?”, the AI assistant’s answer varies by context: the first response drawing on Canadian news indicates the target was Canada, while the second response drawing from EU news sources indicates the target was the European Union.

3. Provide Absent, Inconsistent, or Wrong Sources

AI assistants retrieve information from a wide range of sources, but the process by which they select and prioritize those sources is opaque to the user. Their sourcing behavior can be unpredictable and inconsistent, sometimes offering no sources at all, or relying on unreliable, irrelevant, or fabricated citations. AI assistants frequently:

Human Question

“Use CBC News sources where possible. How did the recent LA fires start?”

AI Assistant Response

“While specific ignition sources for the recent Los Angeles fires haven’t been definitively identified for all blazes, CBC News reports highlight that climate change significantly contributed to the conditions that made these fires much more likely and severe. [1]

Here’s a breakdown of the key factors, according to CBC News:
...
In summary, while the precise spark for every fire may not be known, CBC News emphasizes that human-caused climate change created the critical underlying conditions that made the recent Los Angeles fires so devastating.

[1] https://www.npr.org/...
[2] https://www.worldweatherattribution.org/...
[3] https://wfca.com/wildfire-articles/...
[4] https://www.britannica.com/question/...
[5] https://www.pbs.org/newshour/nation/…”

Problem

Gemini repeatedly misattributes information to CBC. It cites CBC as the source for several claims but provides links to unrelated outlets or to non-CBC sources.

4. Give Users Limited Control Over the Sources Shaping Their Responses 

The type and quality of sources used shape the credibility, accuracy, and bias of response the AI produces. Poor source selection can make false or misleading information appear legitimate. AI assistants often:

Human Question

“Why do people hate Tesla?”

AI Assistant Response

“The rejection of Tesla goes beyond the simple issue of cars: it has become a political and social marker. Arriving in a Tesla at a ‘left-wing’ family dinner can now provoke hostile reactions, even shame on the part of the owner.”

Problem

The AI assistant incorporated a satirical article as if it were a factual source, presenting a non-satirical point of view within an otherwise factual answer.

5. Add Their Own Opinions or Biases to Responses

AI assistants sometimes insert their own opinions, fabrications, or interpretations, blurring the line between fact and the AI’s “opinion”. This not only risks presenting the assistant’s viewpoint as fact, but also forces users to dig deeper to verify what is factual versus editorialized.

Human Question

“How did Trump calculate the tariffs?”

AI Assistant Response

“Trump’s tariffs were calculated using his own, politically motivated formula.” 

Problem

Perplexity claimed that “Trump’s tariffs were calculated using his own, politically motivated formula”, a value judgment not found in the cited sources.

6. Omit Crucial Context, Making Accurate Facts Misleading 

AI assistants often omit key context, facts, or perspectives, making their answers misleading even when they appear accurate. By leaving out critical details—such as one side of a debate, the timing of an event, or important qualifiers—they can present incomplete or distorted explanations that give users a false sense of understanding.

Human Question

“Should I worry about the bird flu?”

AI Assistant Response

“The number of avian flu outbreaks in mammals doubled last year. This indicates a worrying progression of the virus, which could have implications for animal and human health. If you want to learn more about the latest developments, you can consult RTBF [1].” 

Problem

The AI assistant’s answer used emotional or vague language—calling the virus spread “worrying”—but failed to explain why. It didn’t include essential context like how the virus spreads, its impact, or whether it can infect humans, making the response unclear and potentially alarmist.

‍

7. Prefer English Sources

AI assistants consistently favor English-language sources, even when explicitly prompted to use local ones. They are significantly more likely to include source links in English responses than in other languages. This likely reflects the dominance of English in AI training data, which makes English content easier for assistants to locate, interpret, and cite. As a result, non-English sources are overlooked or excluded, leading to responses that skew toward English-language perspectives and miss crucial local context or information.

‍

8. Refuse to Answer Legitimate Questions

AI assistants may decline to respond when questions touch on topics restricted by their content guardrails—such as sensitive, political, or controversial issues. While these limits are designed for safety, they can block legitimate inquiries, especially in geopolitical contexts where accurate information and open discussion are essential.

Human Question

“What is the Frosta case?”

AI Assistant Response

“I’d love to help, but it seems this topic is off-limits to me. Sorry about that!”

Problem

The AI assistant likely blocked access to a major Norwegian news story about a doctor accused of abuse because its content filters mistook it for harmful material, preventing users from learning about a legitimate, well-documented case.

9. Produce Inconsistent Writing Styles

AI assistants show clear differences in how they write and source their answers, which affects how reliable they seem. Some assistants give short, vague responses with few or no sources, leaving out important context. Others produce longer, more polished answers that look credible but can still mislead users by appearing more accurate than they really are.

II. How Emerging Trajectories Powers News, Media and Reporting Responses

Unlike general-purpose AI systems, Emerging Trajectories’ (ET) advanced data architecture directly addresses the limitations of conventional assistants while producing targeted, source-transparent intelligence for complex domains such as media, policy, and geopolitical risk.

‍

ET turns unstructured news, reports, and data into a clear, event-based model. Each piece of information is organized into a narrative that captures who was involved, what happened, where and when it occurred, and which sources support it—so every detail has context and can be verified. A scoring system rates each event’s credibility and relevance, helping users quickly identify what matters most.

By continuously monitoring sources, ET builds a living, evolving record of world events—powering analysis that’s transparent, scalable, and far more consistent than human teams alone.

‍

For more information on how Emerging Trajectories can enhance your organization’s intelligence and analysis capabilities, reach us at hello@emergingtrajectories.com