AI in Journalism: How Indianapolis Newsrooms Use ChatGPT, Otter & NotebookLM — And What Readers Should Demand

Black Indy Live: The Most Experimental Approach

Black Indy Live has leaned furthest into AI. Editor-in-chief Laron Anderson has used it to design logos, generate graphics, polish stories, and build promotional campaigns. He even created an AI avatar named Rae to report the news when the platform launched in 2017.

Anderson’s logic is strategic: AI levels the playing field against larger, better-resourced competitors. Rae was eventually discontinued only because OpenAI shut down Sora, the platform that powered the avatar — not because of editorial concerns.

This is the most aggressive use of AI among Indianapolis outlets, and it raises the most ethical questions.

The Conservative Majority: Transcription First

Most Indianapolis newsrooms are using AI in a much narrower, lower-risk way — primarily for transcription. Here’s the breakdown:

• WFYI uses AI for transcriptions and captions, operating under a “person-first/person-last” policy. A real journalist starts the reporting and reviews the final output before publication.
• Indiana Capital Chronicle uses AI to transcribe recordings, with reporters personally verifying every quote.
• Mirror Indy uses AI to transcribe notes and is actively developing a broader policy, including community input on what readers want to know.
• Chalkbeat Indiana uses AI to transcribe some meetings while still attending key ones in person. Internal guidelines are in place; a public-facing statement is in development.
• Axios Indianapolis has a formal partnership with OpenAI’s ChatGPT, which reporters use to analyze data and sort through documents — a more advanced application than simple transcription.
• IndyStar requires journalists to follow strict company guidelines emphasizing transparency and human oversight.
• FOX59 and CBS4, as Nexstar properties, have a clear policy: all writing and reporting must be the work of human journalists. AI cannot replace human judgment or critical thinking, and any AI use must be disclosed to the audience.
• Indianapolis Business Journal does not have a publicly available AI policy — a gap worth noting.

The Tools Gaining the Most Traction

According to Alex Mahadevan, faculty member at the Poynter Institute and director of MediaWise, the most widely adopted AI tools in journalism right now are transcription tools like Otter. They’re practical, low-risk, and save significant time.

Chatbots like ChatGPT, Claude, and Google’s Gemini are gaining ground for research and data analysis. And Google’s NotebookLM is emerging as a newsroom favorite for organizing notes, interviews, and documents on individual stories — a use case confirmed by Benjamin Toff, associate professor at the University of Minnesota and former senior research fellow at the Reuters Institute for the Study of Journalism at Oxford.

The Non-Negotiables

Both experts point to the same core principle: keep humans in the loop.

Chalkbeat’s Bureau Chief MJ Slaby put it plainly:

“It’s never acceptable to ask AI to write a story and pass it off as your own or use AI to create or alter images.”

Mirror Indy’s managing editor of innovation, Amanda Kingsbury, was equally direct:

“AI will not replace our journalists. We don’t use generative AI to write full articles or portions of articles. We don’t use AI to produce or alter photos, videos, illustrations, or graphics.”

These aren’t just reassuring statements. They’re the baseline standard readers should expect from any credible newsroom.

Look for Specific, Meaningful Disclosures

Vague disclosures are nearly useless. A disclosure that says “AI was used in the production of this content” tells you almost nothing. A good disclosure explains exactly what AI did — whether it reformatted a broadcast story for the web, transcribed an interview, or translated content into another language.

WRTV and WISH-TV are cited as examples of outlets doing this reasonably well. WRTV’s disclosures reference the editorial team and adherence to journalistic standards. WISH-TV explains how content was originally created and then reformatted by AI for digital publication. That level of specificity is what you should expect.

Verify That Humans Are Still Doing the Core Work

The “person-first/person-last” model used by WFYI is a practical standard worth looking for. News Director Sarah Neal-Estes explains it simply:

“We don’t just let AI do the work, we use it as a tool. For example, I do the interview first, person-first. I put it in Otter, then I fix that transcript, person-last; OtterAI always has mistakes.”

That acknowledgment — that AI tools make mistakes and humans catch them — is exactly the kind of honest, grounded approach that builds trust.

Don’t Chase Volume — Vet Your Sources

AI makes it easy to produce enormous quantities of content. That means more noise, more duplication, and more low-quality stories that look credible on the surface.

Benjamin Toff’s advice is practical: don’t read 15 versions of the same story. Instead, identify two or three outlets you trust and understand why you trust them. Look for organizations investing in original reporting, confirming information independently, and doing the hard work that distinguishes journalism from content generation.

“At a minimum, I think people ought to be looking at the nature of the organization that they’re getting information from,” Toff said. “There are still some very good, in my view, quality news organizations that are investing in original reporting and news gathering, confirming information and doing the hard work that is journalism.”

Recognize When AI Expansion Is Actually Good

Not every AI application in journalism is a red flag. When AI translates stories into different languages, converts text to audio for accessibility, or helps reporters sift through thousands of public documents for investigative work — that’s AI serving the public interest.

WFYI regularly provides audio versions of written stories. Mirror Indy is exploring the same capability. The New York Times and The Washington Post already do it. When AI expands access without replacing human judgment, it’s a net positive for readers.