We're Using AI for Answers, But We Don't Trust It For News. Here's Why.
A new Reuters Institute report shows that people are increasingly using AI for quick information but remain deeply skeptical of it creating their news.
The short version
A new report confirms what many of us feel in our guts. We’re using tools like ChatGPT for quick answers, but we draw a hard line at letting AI write the news we read. This is a story about a tool for information, not a source of truth.
What did this new report actually find?
I’ve been watching the public’s relationship with artificial intelligence evolve, and frankly, it feels like we’re all collectively holding two very different ideas in our heads at the same time. On one hand, the convenience is undeniable. On the other, there’s a deep-seated hesitation. A new report on generative AI and news from the Reuters Institute just put hard numbers on that exact feeling. The report surveyed people across nine countries and found a fascinating split.
The headline number is this: 38% of people had used a generative AI tool to get information in the past week. Think about that. More than a third of us are already integrating these tools into our basic information-gathering routines. We’re asking for summaries, definitions, and quick facts. That is a significant adoption rate for a technology that, for most people, is barely a couple of years old. The use is real and it’s growing quickly.
But here’s the other side of the coin. The same report found widespread discomfort with the idea of news being primarily created by AI, especially for sensitive topics like politics or crime. People are drawing a clear, bright line. Using an AI to get a quick summary of a historical event is one thing. Trusting an AI to report on this morning’s political developments is something else entirely. The report shows we are happy to use AI as a super-powered search engine, but we are not ready to call it a journalist. That distinction is everything.
Why are we using AI if we don’t trust it?
This is the question at the heart of the report’s findings. The answer appears to be simple and deeply human: speed and convenience. It’s about efficiency. The AI is a shortcut.
I see this in my own workflow every single day. Just last week, I needed to understand the core components of a new programming framework. My options were to read the dense official documentation, hunt through several blog posts of varying quality, or just ask an AI. I asked Perplexity. I got a structured, bullet-point answer in about ten seconds. Was it a perfect, nuanced explanation? No. Did it give me the vocabulary I needed to then go read the documentation intelligently? Absolutely. It saved me half an hour of searching. That’s its job.
This is information retrieval, not deep analysis. It’s the same reason I might ask an AI for a dinner recipe using the ingredients I have in my fridge or to draft a polite-sounding email declining an invitation. These are low-stakes tasks. The cost of the AI getting it slightly wrong is minimal. If the recipe suggests too much salt, I can adjust. If the email sounds a bit robotic, I can edit it. No real harm done.
News is a different category of information. The stakes are much, much higher. A factual error in a news report about an election, a public health crisis, or a court case can have significant consequences. Trust in journalism is built on a foundation of verification, accountability, sourcing, and ethical judgment. When a human journalist gets something wrong, there’s a process. There are corrections, editor’s notes, and a byline attached to a real person whose reputation is on the line. With an AI, that chain of accountability is murky at best. Who is responsible for a hallucination presented as fact? The news outlet? The company that built the large language model? This ambiguity is a massive barrier to trust.
What does this ‘distrust’ of AI in news actually look like?
The skepticism identified in the Reuters Institute report is not just a vague feeling; it’s rooted in specific, tangible concerns. These worries have been fueled by high-profile failures that show exactly what can go wrong when automation is applied carelessly to the creation of content we are meant to trust.
First, there’s the problem of accuracy and “hallucinations.” AI models are designed to generate plausible text, not to verify facts. They can, and often do, invent information, sources, and events with complete confidence. We saw this with the early AI-written articles from CNET, which were later found to contain numerous simple errors and instances of plagiarism. Readers noticed. The damage to CNET’s reputation was immediate. It was a clear signal that simply outsourcing article creation to a machine without rigorous human oversight was a recipe for disaster.
Then there is the issue of fake identities and sourcing. The scandal at Sports Illustrated, where the publication was found to be publishing articles under the names of fake, AI-generated authors with AI-generated headshots, was a profound breach of trust. The byline on an article is a promise. It’s a promise that a real person did the work, stands by it, and is accountable for it. Creating fake people to attach to AI-generated content obliterates that promise. It makes the reader question everything else the publication produces. You start to wonder what else is fake.
Finally, there’s the potent threat of visual disinformation. It’s one thing to read a poorly written AI article, but it’s another entirely to see a photorealistic image or video of an event that never happened. As AI image and video generation tools become more sophisticated and accessible, the potential for them to be used to create compelling “evidence” of fake events is enormous. An AI-generated image of a politician in a compromising situation or a fake video of a public official giving an inflammatory speech can spread like wildfire online before it can be debunked. This concern about a tsunami of believable fakes is a major reason for the public’s hesitancy to trust AI in a news context.
So, what should news organizations do?
The report isn’t telling newsrooms to abandon AI. That would be like telling a carpenter to abandon a power saw. Instead, the findings offer a clear path for how to integrate this technology responsibly, in a way that respects the audience’s valid skepticism. The solution is about making the AI a tool for the journalist, not a replacement for the journalist.
Transparency is the first and most important principle. If an article, summary, or graphic is created with the assistance of AI, it needs to be clearly and prominently labeled. No exceptions. A simple note like, “This article was summarized by an AI assistant and reviewed by a human editor,” builds trust. It tells the reader that the organization is not trying to hide anything and that a human is still in the loop and accountable for the final product. Hiding the use of AI is a short-term trick that will lead to long-term ruin for a news brand.
Second, newsrooms should focus on using AI for assistive tasks that empower human reporters. The technology is brilliant at things journalists find tedious. Transcribing a two-hour interview. Analyzing a 500-page city budget proposal to flag major spending changes. Searching through decades of archives for related stories. These are tasks where AI can save hundreds of hours, freeing up journalists to do the work that only humans can: building relationships with sources, asking tough questions, investigating leads, and providing nuanced analysis.
Imagine a small local newsroom. They don’t have a large staff. A developer proposes a massive new construction project, submitting thousands of pages of environmental impact reports. A single reporter could spend weeks just trying to read it all. But an AI could be tasked to scan those documents in minutes, identify all mentions of specific chemical compounds, flag any inconsistencies with previous reports, and summarize the key findings. The AI does the grunt work. The reporter then takes that information, calls experts, interviews residents, and writes the story. That’s the smart way to use this technology. It makes the journalism better, not cheaper.
What does this mean for me, the reader?
For those of us who read the news, this report is a validation. Your skepticism is not only warranted; it’s healthy. It means you understand the difference between raw information and verified journalism. Keep that critical eye.
Continue to use AI tools for what they’re good at. They are amazing for brainstorming, summarizing non-critical documents, writing bits of code, or just satisfying your curiosity about a random topic. They are powerful assistants. That’s fine. Go for it.
But for the information you use to make decisions—about your health, your finances, your community, and your government—your standards must be higher. Always ask where your news is coming from. Is there a real author with a byline? Does the publication have a public corrections policy? Does it have a track record of accountability? These are the signals of trust that no AI can replicate.
This is also a moment to consider the value of supporting the news organizations that invest in human beings. Real reporting is expensive. It involves sending people to places, spending time digging through records, and protecting sources. The convenience of an AI chatbot is seductive, but it doesn’t do that work. By paying for a subscription to a reputable news source, you are directly funding the human judgment, ethics, and accountability that this report shows we all still crave. The data is clear: we know there’s a difference between a quick answer and the actual news.
FAQ
What is the main finding of the Reuters Institute report? The report’s central finding is that while people are rapidly adopting generative AI for quick, convenient information tasks, they remain deeply skeptical and uncomfortable with news content being created primarily by AI.
Why are people using AI if they don’t trust it for news? The primary reasons are speed and convenience. AI tools excel at providing fast summaries, answering specific questions, and assisting with low-stakes tasks, which is a fundamentally different use case from consuming vetted journalism.
Is it bad for news organizations to use AI? Not necessarily. The report suggests a responsible path forward where newsrooms use AI as a tool to assist human journalists—for tasks like data analysis or transcription—rather than replacing them. Transparency with the audience about when and how AI is used is crucial.
How can I tell if a news article was written by AI? Reputable news organizations should have clear policies and labels indicating AI assistance. If there is no label, be wary of articles with no author byline, generic or error-filled text, and a lack of original sourcing. The best approach is to rely on publications with established editorial standards and a history of accountability.