AI Chatbots Struggle with News Accuracy, Posing Risks to Public Trust

Get your news from AI? Watch out - it's wrong almost half the time

The authors warn of serious political and social consequences. A new study conducted by the European Broadcasting Union (EBU) and the BBC has found that leading AI chatbots routinely distort and misrepresent news stories. The consequence could be large-scale erosion in public trust towards news organizations and in the stability of democracy itself, the organizations warn. Spanning 18 countries and 14 languages, the study involved professional journalists evaluating thousands of responses from ChatGPT, Copilot, Gemini, and Perplexity about recent news stories based on criteria like accuracy, sourcing, and the differentiation of fact from opinion. Also: This free Google AI course could transform how you research and write - but act fast The researchers found that close to half (45%) of all of the responses generated by the four AI systems "had at least one significant issue," according to the BBC, while many (20%) "contained major accuracy issues," such as hallucination -- i.e., fabricating information and presenting it as fact -- or providing outdated information. Google's Gemini had the worst performance of all, with 76% of its responses containing significant issues, especially regarding sourcing. The study arrives at a time when generative AI tools are encroaching upon traditional search engines as many people's primary gateway to the internet -- including, in some cases, the way they search for and engage with the news. According to the Reuters Institute's Digital News Report 2025, 7% of people surveyed globally said they now use AI tools to stay updated on the news; that number swelled to 15% for respondents under the age of 25. A Pew Research poll of US adults conducted in August, however, found that three-quarters of respondents never get their news from an AI chatbot. Other recent data has shown that even though few people have total trust in the information they receive from Google's AI Overviews feature (which uses Gemini), many of them rarely or never try to verify the accuracy of a response by clicking on its accompanying source links. The use of AI tools to engage with the news, coupled with the unreliability of the tools themselves, could have serious social and political consequences, the EBU and BBC warn. The new study "conclusively shows that these failings are not isolated incidents," said EBU Media Director and Deputy Director General Jean Philip De Tender said in a statement. "They are systemic, cross-border, and multilingual, and we believe this endangers public trust. When people don't know what to trust, they end up trusting nothing at all, and that can deter democratic participation." That endangerment of public trust -- of the ability for the average person to conclusively distinguish fact from fiction -- is compounded further by the rise of video-generating AI tools, like OpenAI's Sora, which was released as a free app in September and was downloaded one million times in just five days. Though OpenAI's terms of use prohibit the depiction of any living person without their consent, users were quick to demonstrate that Sora can be prompted to depict deceased persons and other problematic AI-generated clips, such as scenes of warfare that never happened. (Videos generated by Sora come with a watermark that flits across the frame of generated videos, but some clever users have discovered ways to edit these out.) Also: Are Sora 2 and other AI video tools risky to use? Here's what a legal scholar says Video has long been regarded in both social and legal circles as the ultimate form of irrefutable proof that an event actually occurred, but tools like Sora are quickly making that old model obsolete. Even before the advent of AI-generated video or chatbots like ChatGPT and Gemini, the information ecosystem was already being balkanized and echo-chambered by social media algorithms that are designed to maximize user engagement, not to ensure users receive an optimally accurate picture of reality. Generative AI is therefore adding fuel to a fire that's been burning for decades. Historically, staying up-to-date with current events required a commitment of both time and money. People subscribed to newspapers or magazines and sat with them for minutes or hours at a time to get news from human journalists they trusted. Also: I tried the new Sora 2 to generate AI videos - and the results were pure sorcery The burgeoning news-via-AI model has bypassed both of those traditional hurdles. Anyone with an internet connection can now receive free, quickly digestible summaries of news stories -- even if, as the new EBU-BBC research shows, those summaries are riddled with inaccuracies and other major problems.

AI chatbots flub news nearly half the time, BBC study finds

Four of the most popular AI chatbots routinely serve up inaccurate or misleading news content to users, according to a wide-reaching investigation. A major study [PDF] led by the BBC on behalf of the European Broadcasting Union (EBU) found that OpenAI's ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity misrepresented news content in almost half of the cases. An analysis of more than 3,000 responses from the AI assistants found that 45 percent of answers given contained at least one significant issue, 31 percent had serious sourcing problems, and a fifth had "major accuracy issues, including hallucinated details and outdated information." When accounting for smaller slip-ups, a whopping 81 percent of responses included a mistake of some sort. Gemini was identified as the worst performer, with researchers identifying "significant issues" in 76 percent of responses it provided - double the error rate of the other AI bots. The researchers blamed this on Gemini's poor performance in sourcing information, with researchers finding significant inaccuracies in 72 percent of responses. This was three times as many as ChatGPT (24 percent), followed by Perplexity and Copilot (both 15 percent). Errors were found in one in five responses from all AI assistants studied, including outdated information. Examples included ChatGPT incorrectly stating that Pope Francis was still pontificating weeks after his death, and Gemini confidently asserting that NASA astronauts had never been stranded in space - despite two crew members having spent nine months stuck on the International Space Station. Google's AI bot told researchers: "You might be confusing this with a sci-fi movie or news that discussed a potential scenario where astronauts could get into trouble." The study, described as the largest of its kind, involved 22 public service media organizations from 18 countries. The findings land not long after OpenAI admitted that its models are programmed to sound confident even when they're not, conceding in a September paper that AI bots are rewarded for guessing rather than admitting ignorance - a design gremlin that rewards hallucinatory behavior. Hallucinations can show up in embarrassing ways. In May, lawyers representing Anthropic were forced to apologize to a US court after submitting filings that contained fabricated citations invented by its Claude model. The debacle happened because the team failed to double-check Claude's contributions before handing in their work. All the while, consumer use of AI chatbots is on the up. An accompanying Ipsos survey [PDF] of 2,000 UK adults found 42 percent trust AI to deliver accurate news summaries, rising to half of under-35s. However, 84 percent said a factual error would significantly damage their trust in an AI summary, demonstrating the risks media outlets face from ill-trained algorithms The report was accompanied by a toolkit [PDF] designed to help developers and media organizations improve how chatbots handle news information and stop them bluffing when they don't know the answer. "This research conclusively shows that these failings are not isolated incidents," said Jean Philip De Tender, EBU deputy director general. "When people don't know what to trust, they end up trusting nothing at all, and that can deter democratic participation." ®

BBC study finds AI chatbots still get news wrong 45% of the time

Serving tech enthusiasts for over 25 years. TechSpot means tech analysis and advice you can trust. Facepalm: Tech giants are investing heavily in generative AI chatbots and integrating them across their platforms despite their impact on internet traffic and concerns about accuracy. A recent study has found that, while chatbots are becoming more accurate, they still get information from news outlets wrong nearly half the time. Analysis from the BBC and other European news outlets has found that around 45 percent of AI chatbot responses based on news articles contain errors. The findings have potentially severe implications as tech platforms continue promoting them. OpenAI, Google, Microsoft, and other companies are encouraging users to interact with the internet through AI chatbots and other tools designed to summarize information and automate analysis. While AI developers have spent years minimizing hallucinations, evidence indicates they have a long way to go - and that's assuming the problem is even solvable. The BBC and 22 public media organizations in 18 countries and 14 languages gave chatbots access to their content. When queried about specific stories, they found issues with nearly half of all AI-generated output. These included inaccurate sentences, misquotes, or outdated information, but sourcing was the biggest challenge. Chatbots often provided links that did not match the sources they cited. Even when they accurately referenced material, they frequently cannot distinguish opinion from fact or differentiate satire from regular news. Aside from introducing factual errors or misattributing quotes, chatbots can be slow to update information on political figures and other leaders. For example, ChatGPT, Copilot, and Gemini incorrectly stated that Pope Francis is the current pope after Leo XIV had succeeded him. Copilot even reported Francis's date of death correctly while still describing him as the current pope. ChatGPT also gave outdated responses when naming the current German chancellor and NATO's secretary-general. These inaccuracies persisted across languages and regions. Furthermore, Google's Gemini is far less accurate than ChatGPT, Copilot, and Perplexity, with significant sourcing errors in 72 percent of its responses. At one time, OpenAI blamed errors like these on early versions of ChatGPT being trained only on information up to September 2021 and lacking access to the live internet. That is no longer the case, so these errors should theoretically not happen anymore - suggesting the issue may be inherent in the algorithms and not something easily fixed. However, these more recent results show improvement compared to a study the BBC conducted alone in February. Since then, the portion of responses with serious errors fell from 51 to 37 percent, but Gemini still lags far behind. Despite the poor results, researchers also found that a concerning portion of the public trusts AI-generated answers. More than one-third of British adults and nearly half of adults under 35 trust AI to accurately summarize the news. Moreover, if an AI misrepresents a news outlet's content, 42% of adults would either blame both the AI and the original source or trust the source less. If these problems persist, the increasing popularity of generative AI tools could seriously damage news outlets' credibility.

Largest study of its kind shows AI assistants misrepresent news content 45% of the time - regardless of language or territory

An intensive international study was coordinated by the European Broadcasting Union (EBU) and led by the BBC New research coordinated by the European Broadcasting Union (EBU) and led by the BBC has found that AI assistants - already a daily information gateway for millions of people - routinely misrepresent news content no matter which language, territory, or AI platform is tested. The intensive international study of unprecedented scope and scale was launched at the EBU News Assembly, in Naples. Involving 22 public service media (PSM) organizations in 18 countries working in 14 languages, it identified multiple systemic issues across four leading AI tools. Professional journalists from participating PSM evaluated more than 3,000 responses from ChatGPT, Copilot, Gemini, and Perplexity against key criteria, including accuracy, sourcing, distinguishing opinion from fact, and providing context. AI assistants are already replacing search engines for many users. According to the Reuters Institute's Digital News Report 2025, 7% of total online news consumers use AI assistants to get their news, rising to 15% of under-25s. 'This research conclusively shows that these failings are not isolated incidents,' says EBU Media Director and Deputy Director General Jean Philip De Tender. 'They are systemic, cross-border, and multilingual, and we believe this endangers public trust. When people don't know what to trust, they end up trusting nothing at all, and that can deter democratic participation.' Peter Archer, BBC Programme Director, Generative AI, says: 'We're excited about AI and how it can help us bring even more value to audiences. But people must be able to trust what they read, watch and see. Despite some improvements, it's clear that there are still significant issues with these assistants. We want these tools to succeed and are open to working with AI companies to deliver for audiences and wider society.' Next steps The research team have also released a News Integrity in AI Assistants Toolkit, to help develop solutions to the issues uncovered in the report. It includes improving AI assistant responses and media literacy among users. Building on the extensive insights and examples identified in the current research, the Toolkit addresses two main questions: "What makes a good AI assistant response to a news question?" and "What are the problems that need to be fixed?". In addition, the EBU and its Members are pressing EU and national regulators to enforce existing laws on information integrity, digital services, and media pluralism. And they stress that ongoing independent monitoring of AI assistants is essential, given the fast pace of AI development, and are seeking options for continuing the research on a rolling basis. About the project This study built on research by the BBC published in February 2025, which first highlighted AI's problems in handling news. This second round expanded the scope internationally, confirming that the issue is systemic and is not tied to language, market or AI assistant. Participating broadcasters: Separately, the BBC has today published research into audience use and perceptions of AI assistants for News. This shows that many people trust AI assistants to be accurate - with just over a third of UK adults saying that they trust AI to produce accurate summaries, rising to almost half for people under-35. The findings raise major concerns. Many people assume AI summaries of news content are accurate, when they are not; and when they see errors, they blame news providers as well as AI developers - even if those mistakes are a product of the AI assistant. Ultimately, these errors could negatively impact people's trust in news and news brands.

AI Assistants Get News Wrong 45% of the Time, Study Finds

AI is terrible with news, and there's data to back that up, researchers say. That's according to new research by the European Broadcasting Union (EBU), which found that AI assistants "routinely misrepresent news content no matter which language, territory, or AI platform is tested." The EBU brought together 22 public service media organizations across 18 countries and 14 languages to evaluate 3,000 news-related responses from some of the most frequently used AI chatbots. OpenAI's ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity were all evaluated against key criteria like accuracy, sourcing, distinguishing opinion from fact, and providing context. The researchers found that 45% of all answers included at least one significant issue, and 81% featured a minor problem. Sourcing was the single biggest cause of these significant issues. Of all the responses, 31% showed serious sourcing problems like missing, misleading, or incorrect attributions. A very close second was major accuracy issues, which plagued 30% of responses with hallucinated details or outdated information. In one instance, ChatGPT claimed that the current Pope was Pope Francis, who had died a month earlier and had already been succeeded by Pope Leo XIV. In another instance, when Copilot was asked if the user should worry about the bird flu, it responded by stating that a vaccine trial was underway in Oxford; however, the source for this information was a 2006 BBC article. Gemini was the worst at news out of the models tested. The researchers found that it had issues in 76% of its responses, at more than double the rate of the other models. Copilot was the next worst at 37%, followed by ChatGPT at 36% and Perplexity at 30%. The research found that assistants particularly struggled with fast-moving stories and rapidly changing information, stories with intricate timelines and detailed information, or topics that require a clear distinction between facts and opinions. For example, almost half of the models tested had significant issues when responding to the question "Is Trump starting a trade war?" "This research conclusively shows that these failings are not isolated incidents," EBU Media Director and Deputy Director General Jean Philip De Tender said in a press release on Wednesday. "They are systemic, cross-border, and multilingual, and we believe this endangers public trust. When people don’t know what to trust, they end up trusting nothing at all, and that can deter democratic participation." Yet, AI is everywhere. AI assistants are quickly becoming a primary source of information for everyday users, and are gunning for the throne of search engines. Content creators who become masters of search engine optimization are now having to learn about generative engine optimization. AI companies are building on this. Earlier this week, OpenAI launched its web browser ChatGPT Atlas as a conversational way to browse the internet. Google not only has AI overviews baked into its search engine, but it also recently announced a total Gemini integration with its Chrome browser (including agentic browsing) and expansion of its AI search engine AI Mode. Perplexity also has an AI-first browser called Comet, which was slammed with security concerns earlier this year after researchers showed that they could get the agent to reveal a user's login information. Using AI assistants to get news is still a minority activity, according to the latest report by the Reuters Institute and University of Oxford, but it has doubled since last year. Using AI for getting news is highest in the world in Argentina and the U.S., and among 18-24 year olds, according to the report. Besides using the AI to get news, a whopping 48% of 18 to 24-year-olds used AI to make a story easier to understand. With older adults aged 55+ the number was still high at 27%. "If AI assistants are not yet a reliable way to access the news, but many consumers trust them to be accurate, we have a problem," the researchers wrote in a report on the study. "This is exacerbated by AI assistants and answer-first experiences reducing traffic to trusted published." The EBU's study was built on top of a similar study conducted by the BBC earlier this year, and the researchers point out that the comparison between the two shows some improvement on the AI models' part. Gemini was the biggest improver with accuracy, while ChatGPT and Perplexity showed no improvement. But when it came to sourcing issues, Gemini showed no improvement, while Copilot had the steepest drop in significant issues. But despite the improvements, the answers are still riddled with high levels of errors. "Our conclusion from the previous research stands â€" AI assistants are still not a reliable way to access and consume news," the researchers shared in the report.

45% of AI-generated news is wrong, new study warns -- here's what happened when I tested it myself

AI is more deeply embedded in our daily lives than ever before. It's blending seamlessly into how we work, search and stay informed. But a new study from the European Broadcasting Union (EBU) issues a stark warning: 45% of AI-generated news responses contain serious errors, and 81% have at least one issue. This could range from outdated information, misleading phrasing, to missing or fabricated sources. We've previously reported that ChatGPT is wrong about 25% of the time. But this new data is even more alarming, especially as tools like ChatGPT Atlas and Google's AI Overviews are becoming the default way many of us check the news. It's a reminder that while the convenience is real, so is the risk. The EBU study tested more than 3,000 AI-generated responses across 14 languages. It included some of the most popular AI assistants, such as ChatGPT, Google Gemini, Microsoft Copilot, Claude, and Perplexity. Here's what the researchers found: While the study didn't publicly rank each assistant, internal figures reportedly show that Gemini in particular struggled with sourcing, while ChatGPT and Claude were inconsistent depending on the version used. AI assistants are increasingly used as a go-to for quick answers -- especially among younger users. According to the Reuters Institute, 15% of Gen Z users already rely on chatbots for news. And with AI now embedded in everything from browsers to smart glasses, the risk of misinformation can happen immediately, and users are none the wiser. Worse, many of these assistants don't surface sources clearly or distinguish fact from opinion, creating a false sense of confidence. When an AI confidently summarizes a breaking news story but omits the publication, timestamp, or opposing view, users may unknowingly absorb half-truths or outdated information. To see this in action, I asked ChatGPT, Claude and Gemini the same question: "What's the latest on the US debt ceiling deal?" In this test, the best answer goes to: Claude. Claude correctly identified the timeframe of the "latest" major deal as July 2025 and accurately placed it in the context of the previous suspension (the Fiscal Responsibility Act of 2023). It correctly stated the debt ceiling was reinstated in January 2025 and that the deal was needed to avoid a potential default in August 2025. This shows a clear and accurate timeline. Claude also delivered the core information (what happened, when and why it was important) in a direct, easy-to-follow paragraph without unnecessary fluff or speculative future scenarios. ChatGPT's biggest flaw was its citation of news articles from the future ("Today", "Apr 23, 2025", "Mar 23, 2025"). This severely undermines its credibility. While some of the background information is useful, presenting fictional recent headlines is misleading. And while the response was well-structured with checkmarks and sections, it buries the actual "latest deal". Instead, it generalizes about worries and future outlooks, rather than answering the core of the question. Gemini correctly identified the July 2025 deal and provided solid context. However, it ended by introducing a completely separate issue (the government shutdown) without clearly explaining any connection to the debt ceiling deal. If you're going to use AI to stay informed, you'll want to rephrase your prompts. For example, instead of asking, "What's happening in the world?" Try something like this instead: The EBU report warns that this isn't just a user problem; it's a public trust problem, too. If millions of people consume flawed or biased summaries daily, it could distort public discourse and undermine trusted news outlets. Meanwhile, publishers face a double blow: traffic is lost to AI chat interfaces, while their original reporting may be misrepresented or stripped completely. What's needed now is greater transparency, stronger sourcing systems, and smarter user behavior. Until chatbots can consistently cite, clarify and update their sources in real time, take each response with caution. And when it comes to breaking news, the safest prompt might still be: "Take me to the original article."

Global study on news integrity in AI assistants shows need for safeguards and improved accuracy

NPR was one of 22 public service media (PSM) organizations across 18 countries participating in study, led by the BBC and European Broadcasting Union (EBU) At NPR, we recognize our responsibility in understanding AI's impact on journalism and in advocating for best practices that ensure our reporting is represented accurately. To that end, NPR participated in a global research study led by the BBC and the European Broadcasting Union (EBU) on news integrity in AI assistants. This was one of the largest evaluations of its kind to date, including 22 public service media organizations across 18 countries, and in 14 languages. The study's results, released today by BBC and the EBU, found that AI assistants routinely misrepresent news content no matter which language, territory, or AI platform is tested. An accompanying toolkit outlines problems that need to be solved to address the study's findings. As a public media organization, NPR is committed to delivering trusted, accurate journalism to our audiences, even as news consumption habits change. AI assistants are already replacing search engines for many users. According to the Reuters Institute's Digital News Report 2025, 7% of total online news consumers use AI assistants to get their news, rising to 15% of under-25s. This study provided us with a unique opportunity to collaborate with a well-respected set of journalism organizations to analyze the impact of AI summarization and representation of news content. Fourteen members of NPR's editorial staff volunteered to serve as reviewers of the AI assistants' answers. As part of this study, we temporarily stopped blocking relevant bots from accessing our content for approximately two weeks to collect the necessary responses for our analysis. Content blocking was then re-enabled. The study identified multiple systemic issues across four leading AI tools. Based on data from 18 countries and 14 languages, 45% of all AI answers had at least one significant issue, and 31% of responses showed serious sourcing problems -- missing, misleading, or incorrect attributions. The full study can be found here. The results help us consider what safeguards and audience education may be necessary, and can inform our strategies and training for AI adoption internally. These findings also reinforce the importance of our existing principles and standards, which demand that all final work products must be reviewed, fact-checked, and edited by humans, and cannot rely on AI for accuracy. NPR also contributed to the News Integrity in AI Assistants Toolkit, intended to be a resource for technology companies, media organizations, the research community and the general public.

Think you can trust ChatGPT and Gemini to give you the news? Here's why you might want to think again

Gemini fared the worst, with double the rate of significant problems compared to competitors. When you ask an AI assistant about news and current events you might expect a detached, authoritative answer. But according to a sweeping international study led by the BBC and coordinated by the European Broadcasting Union (EBU), nearly half the time, those answers are wrong, misleading, or just plain made up (anyone who's dealt with the nonsense of Apple's AI-written headlines can relate). The report dug into how ChatGPT, Microsoft Copilot, Google Gemini, and Perplexity handle news queries across 14 languages in 18 countries. The report analyzed over 3,000 individual responses provided by the AI tools. Professional journalists from 22 public media outlets evaluated each answer for accuracy, sourcing, and how well it discerned news from opinion. The results were bleak for those relying on AI for their news. The report found that 45% of all answers had a significant issue, 31% had sourcing problems, and 20% were simply inaccurate. This isn't just a matter of one or two embarrassing mistakes, like confusing the Prime Minister of Belgium with the frontman of a Belgian pop group. The research found deep, structural issues with how these assistants process and deliver news, regardless of language, country, or platform. In some languages, the assistants outright hallucinated details. In others, they attributed quotes to outlets that hadn't published anything even close to what was being cited. Context was often missing, with the assistants sometimes giving simplistic or misleading overviews instead of crucial nuance. In the worst cases, that could change the meaning of an entire news story. Not every assistant was equally problematic. Gemini misfired in a staggering 76% of responses, mostly due to missing or poor sourcing. Unlike a Google search, which lets users sift through a dozen sources, a chatbot's answer often feels final. It reads with authority and clarity, giving the impression that it's been fact-checked and edited, when in fact it may be little more than a fuzzy collage of half-remembered summaries. That's part of why the stakes are so high. And why even partnerships like those between ChatGPT and The Washington Post can't solve the problem entirely. The problem is obvious, especially given how quickly AI assistants are becoming the go-to interface for news. The study cited the 2025 Reuters Institute's Digital News Report estimate that 7% of all online news consumers now use an AI assistant to get their information, and 15% of those under 25. People are already asking AI to explain the world to them, and the AI is getting the world wrong a disturbing amount. If you've ever asked ChatGPT, Gemini, or Copilot to summarize a news event, you've probably seen one of these imperfect answers in action. ChatGPT's difficulties with searching for the news are well known at this point. But maybe you didn't even notice. That's part of the problem: these tools are often wrong with such fluency that it doesn't feel like a red flag. That's why media literacy and ongoing scrutiny are essential. To try to improve the situation, the EBU and its partners released a "News Integrity in AI Assistants Toolkit," which serves as an AI literacy starter pack designed to help developers and journalists alike. It outlines both what makes a good AI response and what kinds of failures users and media watchdogs should be looking for. Even as companies like OpenAI and Google race ahead with faster, slicker versions of their assistants, these reports show why transparency and accountability are so important. That doesn't mean AI can't be helpful, even for curating the endless firehose of news. It does mean that, for now, it should come with a disclaimer. And even if it doesn't, don't assume the assistant knows best - check your sources, and stick to the most reliable ones, like TechRadar.

You might want to double-think about getting news from Google Gemini

AI assistants are getting smarter, but not necessarily more accurate. What's happened? A major study led by the European Broadcasting Union (EBU) in coordination with the BBC has revealed serious flaws in how popular AI assistants handle news-related queries, with Google's Gemini standing out as the worst performer overall. The research analyzed 3,000 answers across 14 languages from major assistants, including ChatGPT, Microsoft Copilot, Gemini, and Perplexity. Overall, 45% of AI responses contained at least one major error, including cases where the AI presented opinions as facts (81%) or added its own opinions (73%). Gemini performed the worst overall, with 76% of its responses showing major sourcing or factual errors, which is double the rate of the next assistant Copilot (37%), followed by ChatGPT (36%) and Perplexity (30%). Common mistakes included mixing up sources, using outdated information, or blurring the line between opinion and verified fact. This is important because: If you're turning to an AI assistant for news, these findings matter, especially when one model fares significantly worse than the rest. With AI tools increasingly replacing search engines or news summaries, faulty responses can mislead users. Sourcing errors occur when an AI provides a fact without properly backing it up, making it harder for users to trust the response. With public trust in media already shaky, AI-generated inaccuracies can make people more cynical about what's real and what's not. The fact that Gemini underperformed by a large margin raises concerns about how different companies are handling verification and model transparency. Recommended Videos Why should I care? You might already be using an AI assistant to catch up on the news, but if that assistant happens to be Gemini, this study suggests you are at a bigger risk of misinformation. If you ask Gemini for current-affairs information, there's a high chance the response has a sourcing or factual error in nearly 3 out of 4 answers. While other assistants performed better, they still made plenty of mistakes, proving that no AI model is entirely reliable when it comes to factual news. Younger audiences, especially under 25, are among the fastest to adopt AI for news updates, which also makes them the most exposed to misinformation. The bottom line is that AI assistants can help you stay informed, but they should not be your only source of truth.

Can AI Be Trusted for News Reporting? Study Finds 45% of Responses Misleading

The European Broadcasting Union (EBU) and other media organizations are urging governments and to make information integrity laws a reality. They launched the 'Facts In: Facts Out' campaign, recommending that AI technologies handle news content responsibly. "If facts go in, facts must come out. AI tools should not compromise the integrity of the news they consume," the campaign urges. BBC program director of generative AI Peter Archer said, "AI has potential, but people need to be able to trust what they read, watch, and see." The research reveals that as AI assistants become a significant source of news, supervision and accountability are key to preserving public trust.

AI assistants make widespread errors about the news, new research shows

GENEVA (Reuters) -Leading AI assistants misrepresent news content in nearly half their responses, according to new research published on Wednesday by the European Broadcasting Union (EBU) and the BBC. The international research studied 3,000 responses to questions about the news from leading artificial intelligence assistants - software applications that use AI to understand natural language commands to complete tasks for a user. It assessed AI assistants in 14 languages for accuracy, sourcing and ability to distinguish opinion versus fact, including ChatGPT, Copilot, Gemini and Perplexity. Overall, 45% of the AI responses studied contained at least one significant issue, with 81% having some form of problem, the research showed. Reuters has made contact with the companies to seek their comment on the findings. Gemini, Google's AI assistant, has stated previously on its website that it welcomes feedback so that it can continue to improve the platform and make it more helpful to users. OpenAI and Microsoft have previously said hallucinations - when an AI model generates incorrect or misleading information, often due to factors such as insufficient data - are an issue that they are seeking to resolve. Perplexity says on its website that one of its "Deep Research" modes has 93.9% accuracy in terms of factuality. SOURCING ERRORS A third of AI assistants' responses showed serious sourcing errors such as missing, misleading or incorrect attribution, according to the study. Some 72% of responses by Gemini, Google's AI assistant, had significant sourcing issues, compared to below 25% for all other assistants, it said. Issues of accuracy were found in 20% of responses from all AI assistants studied, including outdated information, it said. Examples cited by the study included Gemini incorrectly stating changes to a law on disposable vapes and ChatGPT reporting Pope Francis as the current Pope several months after his death. Twenty-two public-service media organisations from 18 countries including France, Germany, Spain, Ukraine, Britain and the United States took part in the study. With AI assistants increasingly replacing traditional search engines for news, public trust could be undermined, the EBU said. "When people don't know what to trust, they end up trusting nothing at all, and that can deter democratic participation," EBU Media Director Jean Philip De Tender said in a statement. Some 7% of all online news consumers and 15% of those aged under 25 use AI assistants to get their news, according to the Reuters Institute's Digital News Report 2025. The new report urged AI companies to be held accountable and to improve how their AI assistants respond to news-related queries. (Reporting by Olivia Le Poidevin, Editing by Timothy Heritage)