Thinking for Us: Is AI Changing the Way We Understand News?
By: Kalaeja Minyon Jones
In an era where nearly one in four Americans now use artificial intelligence as a primary gateway to information, the traditional bridge between news publishers and the public is undergoing a radical, and some argue dangerous, transformation.
For decades, search engines functioned as intermediaries, directing traffic to original reporting; however, the newest generative AI search tools are increasingly repackaging that information, without sending readers back to the publishers that produced it, often cutting off the very traffic that helps sustain the news industry.
The danger is not only financial. When AI systems become the main filter through which people encounter the news, it can blur the distinction between verified reporting and machine-generated summaries and reduce the public’s ability to evaluate sources for themselves. In that environment, press freedom is threatened not by direct censorship, but by an erosion of journalism’s visibility, authority and independence.
Yet, journalists may be contributing to this problem themselves.
The “Best Intern” in the Newsroom
Within U.S. newsrooms, AI is already a daily reality, though its role is often more clerical than editorial. Brit Stack, a member of an AI task force at the television station group Morgan Murphy Media, describes AI as a tool for efficiency, helping journalists manage a workload that includes shooting video, writing scripts, and creating social media content.
Morgan Murphy uses a platform called Elvex to build custom assistants that translate complex legal jargon or “sciency” press releases into plain English, saving reporters hours of labor.
“AI is the best intern you’ve ever had, but it’s still just an intern. You’ve got to you still have to check your check their work,” Stack said, quoting a phrase popular with her company’s chief innovation officer. She notes that while AI can draft a web article or an anchor script in seconds, it frequently struggles with nuances like pacing and attribution. She said her newsroom’s rule is that the “first set of hands on a story must be human. It cannot be AI.”
At KSL News Radio in Salt Lake City, News Director Becky Bruce remains skeptical of the technology’s ability to handle high-stakes reporting. During a test to see if AI could write a hypothetical obituary for a local public figure in Utah using AP style, she said the results were surface-level accurate but fundamentally problematic.
“I would have had to go over it with a microscope to spot all of the flaws with what it turned out,” Bruce said, warning that if a journalist or consumer isn’t looking that closely, misinformation can easily slip through.
A Crisis of Confidence: The Problem with Citation
The struggle for accuracy with AI tools is not confined to the newsroom. A comprehensive study by the Tow Center for Digital Journalism at Columbia University recently tested eight prominent generative search tools, including ChatGPT, Perplexity, and Google’s Gemini. The findings were clear: chatbots provided incorrect or “completely incorrect” answers to more than 60% of queries.
The issue isn’t just that the AI is wrong; it is that it is confidently wrong. The study found that most tools presented inaccurate answers with an authoritative tone, rarely using qualifying language like “it appears” or “might.” This “unearned confidence” presents users with a dangerous illusion of reliability. Even premium models like Perplexity Pro or Grok 3, which can cost users up to $40 a month, demonstrated higher error rates than their free counterparts because they were more likely to provide a definitive, incorrect answer rather than decline to answer at all.
Isaac Floyd, the sales and acting general manager of Oxford Media Group, said he experienced this firsthand while using Google’s Gemini to find quirky news stories for his radio show. When he asked for a source for a specific story the AI tool had provided, the chatbot admitted it had fabricated the news story entirely.
“That’s scary because I asked [the chatbot] to pull from this source, this source, and this source to find me three odd stories, and here [it is] creating a whole new imaginary story,” Floyd said. He warns that for those who are not tech-literate, this “AI slop” is often taken at face value, leading to a rise in people “believing in every AI article they see.”
The Erosion of Critical Thinking
The algorithms that power these AI tools can also create echo chambers, reinforcing a user’s existing beliefs and making it harder to encounter opposing viewpoints. Floyd notes that this contributes to a “divided dystopia” where bipartisanship dies because users are only fed one side of a story.
“Most people today are going to be in their echo chambers,” he said, “and echo chambers aren’t healthy for anybody.”
Bruce, however, argues that this passivity pre-dates the AI boom. “I don’t think it’s the AI that’s making people passive,” she said, pointing instead to social media algorithms that prioritize engagement over verification. She said the challenge is that the average person is “too busy” to go out of their way to verify information.
This shift toward AI summaries may also impact the depth of audience understanding. Newsrooms within the Morgan Murphy Media station group include “the brief,” a set of bullet points at the top of articles to help time-pressed readers, but the fear is that many never click through to the full story.
Floyd argues that when users rely solely on summaries, they lose the “meat” of the story and the fine details that reveal the true narrative. “You lose the creative thinking side of yourself,” he warns.
Bruce, who is also a co-chair of the Radio Television Digital News Association’s (RTDNA) ethics committee, also advocates for media literacy training starting in elementary schools, teaching the public to spot the difference between propaganda, opinion, and factual reporting.
The Threat to Press Freedom and Sustainability
AI also poses a structural threat to the freedom of the press by undermining the economic model of journalism. When AI chatbots provide answers without sending users to the source, publishers lose the web traffic and advertising income necessary to pay journalists.
The Tow Center study highlighted a “troubling imbalance” where AI tools sometimes disregard the Robot Exclusion Protocol (robots.txt), a web standard that allows publishers to block crawlers. For example, Perplexity was found to have identified excerpts from National Geographic despite the publisher explicitly disallowing the chatbot’s crawlers. By ignoring these protocols, AI companies take away a publisher’s agency to decide how their work is surfaced or monetized.
Furthermore, even when AI tools do provide citations, they are often broken or misleading. The study found that more than half of the responses from Gemini and Grok 3 cited fabricated or broken URLs. In other instances, chatbots linked to syndicated versions of stories on Yahoo News or AOL rather than the original source, depriving the original publisher of credit and referral traffic.
This shift is a direct concern for press freedom, as the ability of journalists to report independently is undermined when AI strips publishers of the ability to control or profit from their own work.
As artificial intelligence continues to reshape the media landscape, the future of press freedom will depend on how these tools are integrated and regulated. Protecting journalism will require stronger standards for transparency, proper attribution, and respect for publishers’ rights, as well as a renewed emphasis on media literacy among the public. Without these safeguards, AI risks becoming not just a tool that changes how we consume the news, but a force that quietly undermines the independence, sustainability, and democratic function of the press itself.