Public media and nonprofit newsrooms are trying to determine how best to leverage artificial intelligence tools in their work. At WUNC in Chapel Hill, N.C., journalists are interested in AI but are concerned about its potential effects on audience trust.
“It can both help, and I think it can also take away some of the human ingenuity and what have you that goes into figuring out what’s important and what’s going to be important to readers and listeners,” said News Director Brent Wolfe.
National news organizations such as the New York Times are using AI to comb through large datasets and perform other tasks that would be difficult if not impossible for humans to do on their own. Yet newsrooms have also run into pitfalls, including AI-generated hallucinations making their way into print. This happened last year when the Chicago Sun-Times published a summer reading list that featured nonexistent books.
As part of my faculty-supervised independent study this semester at the University of North Carolina at Chapel Hill’s Hussman School of Journalism and Media, I discussed AI and its impact on journalism with leaders at three North Carolina–based media outlets — WUNC, PBS North Carolina and NC Local. They each use AI tools in different ways and to varying extents, navigating a media landscape where AI is becoming increasingly common in newsroom operations.
WUNC
As WUNC’s news director since 2011, Wolfe said he has been present for AI’s introduction “into the conversation.” The station’s experimentation with AI has included using it to develop a weekly news quiz. The newsroom used the AI-powered platform NewsQuiz.ai to review WUNC’s published stories from the past week and generate the quiz. Staffers then reviewed the quiz for accuracy. (The station discontinued the quiz in March due to low engagement.)
Wolfe
WUNC also uses AI built into Adobe Audition to clean up hard-to-understand audio. A section about AI added to WUNC’s ethics statement in April says that the newsroom can use AI to polish and verify reporting but not to generate audio for broadcast.
“Any sound or person talking that you hear on the radio is going to be genuine, although we have left the door open for using AI-generated voices on digital platforms,” Wolfe said.
Wolfe said an internal AI committee comprising staffers from every station department created the policy. The committee is tasked with reviewing any questions about AI that might arise in the future.
The most important stipulation in the policy, Wolfe said, is that a staffer must review and fact-check anything AI generates because preserving audience trust is the top priority. Wolfe said he’s concerned about AI’s potential for spreading misinformation.
“There have been some high-profile cases of plagiarism … and we’ve got editors and fact-checking in place to try and prevent that,” he said. “And so we’re going to do the same thing with AI.”
Looking ahead, Wolfe said, AI could benefit WUNC if tools for cleaning up audio improve and if it can help identify interesting stories that should be shared with reporters.
PBS North Carolina
PBS North Carolina is using AI technology to make workflows more efficient, said Deputy GM Laura Kieler in an email. “For example, if an AI tool can quickly create a caption file for a video or organize our footage with user-friendly tags for easy searchability later on, this would allow our team to spend more time on the editorial aspects of their work,” she said.
Kieler
The station is also using it for a significant archival project — digitizing its 70-year back catalog of TV and video content.
As an affiliate of the University of North Carolina System, which comprises 16 public universities across the state, PBS North Carolina looks to the system for AI guidance. The station also consults PBS’ standards. A 2023 PBS memo about use of generative AI emphasized the importance of accuracy, transparency and inclusiveness.
Though AI can be helpful, Kieler said PBS North Carolina also recognizes that it should be used with care, in part because it can perpetuate biases contained in the data large language models are trained on. Such biases can include stereotyping bias, where AI reinforces harmful stereotypes, and selection bias, when training data doesn’t represent the real-world population.
These stereotypes can go beyond race and gender. An MIT study conducted last year found that chatbots such as ChatGPT sometimes gave less accurate information to users who were less proficient in English, had less formal education or were from outside the U.S.
Kieler said concerns like these are why PBS North Carolina prioritizes its integrity when considering uses of AI.
NC Local
NC Local, a nonprofit newsroom based in Wilmington, uses AI as part of transcription services and for behind-the-scenes tasks such as search engine optimization. Editorial Director Laura Lee, who has a background in public media, said she is also considering how it could be used to gather and sift through large amounts of information.
Lee
One example is identifying local government meetings that people can attend and including them in newsletters, Lee said. “We have a reporter who covers the western part of the state that’s arguably between 15 to 25 counties,” Lee said. “So manually looking up every single meeting of every town council and city council and county commission and whatever board is very labor-intensive.”
NC Local’s AI policy, which Lee created, emphasizes from the start that the news organization is “human-driven.” Lee was in part motivated to develop the policy after taking a training session offered by the Online News Association’s AI in Journalism Initiative. The session was hosted by Lynn Walsh, assistant director of Trusting News, a research and training project that aims to address the problem of declining trust in the news industry.
One of the session’s takeaways, Lee said, was the importance of disclosing the use of AI to audiences. A 2024 Trusting News survey about audiences’ perspectives on newsrooms’ use of AI found that 94% of respondents want journalists to disclose when they use AI. NC Local’s AI policy says the organization will disclose the use of AI tools with a short tagline, a caption or an editor’s note. A guide to free food in the state published last year notes that “AI was used in the completion of some of the data fields (such as address or website) for this project” and links to NC Local’s AI policy.
Lee said she hopes reporters in her newsroom will raise cases that identify AI’s opportunities and limitations so they can discuss them with other newsrooms and people studying AI.
“I think one of the most dangerous things we could do as news leaders is sort of say, ‘Well, here’s the bright-line rule and … that’s the end of the conversation,’ because there are complexities and nuances,” she said.