‘A game-changer for local news’: How AI can enhance the work of independent newsrooms when used responsibly

While the number of Artificial Intelligence tools for journalists has exploded over the last year, AI’s impact on local journalism has been growing for years. And here at LION Publishers, we’ve heard from our members that it can be challenging to figure out how AI might be impacting their work, whether to use certain tools, and what the future might look like.

That’s why we asked Nikita Roy, the founder and host of the Newsroom Robots podcast and Harvard-recognized AI futurist, to join us in our New Entrepreneur Community Slack to talk about it.

Nikita is currently a Knight Fellow at the International Center for Journalists, focusing on AI literacy in the news industry. She leads the AI Journalism Lab and serves on the AI advisory board at the Craig Newmark Graduate School of Journalism at the City University of New York. She also leads the development of the AI Discovery Course at JournalismAI at Polis — the journalism think tank at the London School of Economics.

Nikita recently answered questions about AI and local journalism in an “Ask Me Anything” in the NEC Slack. The questions and answers below have been lightly edited for clarity and length:

What got you interested in studying AI and local journalism?

Nikita: My journey in news started with founding a publication for the Indian diaspora, The NRI Nation (a LION member). I also helped LION research the Canadian local news ecosystem last year. As a news entrepreneur and data scientist myself, I’ve been quite curious about how AI can help us in doing better journalism. I firmly believe that AI is a game-changer for local news. Generative AI, in particular, has lowered the barriers to technological innovation.

What are some examples of local journalists using AI that stand out to you?

Nikita: There are some really exciting AI projects happening in local newsrooms, and some of them are LION members, too: 

• Richland Source’s Lede AI: This tool automatically covers high school sports games across Ohio. Jay Allred shared on the podcast how they’ve been using AI to expand local sports coverage since 2018.
• Local News Now: The founder, Scott Brodbeck, has been using generative AI-powered automation to help streamline tasks such as proofreading articles for typos, summarizing stories, analyzing article tone, and assisting in the creation of sponsored content. They even tried an AI-generated morning newsletter summarizing the previous day’s stories, which made the news in Nieman Lab last year. While that specific project’s wrapped up, it showed some cool possibilities for AI in local news. He shared more in this podcast episode.
• Michigan Radio’s “Minutes” Tool: This was part of AP’s Local News AI initiative,  which helped cover local government meetings. It finds city council meeting videos, generates transcripts, and helps journalists keep tabs on local politics more efficiently. Here’s Ernest Kung from the AP describing how they built this product on the podcast.
• Torontoverse’s TorontoBot: This AI chatbot acts as a municipal budget analyst, helping both journalists and the public understand complex civic financial data. You can chat with it on their Discord channel! Chris Dinn, the publisher, shared how he built the bot on the podcast.

What ethical considerations should journalists make when implementing generative AI in their coverage?

Nikita: In my opinion, transparency and disclosure should be our starting points. We need to be clear about how we’re using AI in our newsrooms, how it is for the audience’s benefit, and what oversight systems we have in place. It’s also crucial to be upfront about the extent of AI’s role in our reporting.

Using AI in journalism isn’t inherently problematic. In fact, it can be a powerful tool when used responsibly. However, when we’re dealing with generative AI, we need to be aware of some key issues. Accuracy is a big one — current AI systems aren’t always spot-on and typically need a human in the loop to ensure reliability. We also need to be vigilant about preventing any unintentional biases that might creep in from these systems.

Accountability is another crucial factor. We need to be clear about who’s responsible for the information generated by AI systems. And privacy is another big issue. We must carefully consider how we’re handling the data that goes into these models. Our sources trust us with their information, and we need to think about the potential consequences when that data goes beyond our direct control.

How do you address AI’s issue with “hallucinating,” when AI will generate a misleading or incorrect response?

Nikita: Always have human oversight for any generative AI tool being used in the news. There are ways to reduce hallucination, but that requires technical expertise. Even with sophisticated prompts and technical methods, generative AI systems can hallucinate. As of right now, hallucinations are a fundamental problem in generative AI models and there is no complete fix for it.

Axios recently had its first-ever layoffs, and in his note announcing the layoffs, the Axios CEO noted that AI is changing the media business. So how should journalists handle this? How do we deal with the very real fear that AI will put people out of a job by replacing journalists with machine-generated content?

Nikita: The fear of AI replacing journalists is valid, but it’s important to recognize that our industry has been operating on a broken business model long before AI came into the picture. AI, in my opinion, cannot replace journalists. The essence of journalism is inherently human — building connections, talking to sources, and uncovering stories. AI can assist with the news production, but it cannot replicate the nuanced, empathetic, and investigative nature of journalism.

But an AI-driven information ecosystem brings with it a new set of challenges, and as an industry, we need to reevaluate our value to our audience. We can’t just provide information or the news anymore — anyone can do that. Our focus should be on providing deeper insights, context, trustworthy analysis, and connection to our community that only experienced journalists can deliver. Embracing AI while maintaining and enhancing the core human elements of journalism is the way forward.

Where should a news entrepreneur start if they want to use AI as a tool in their newsroom, particularly when they are a small operation without a lot of resources?

Nikita: As a news entrepreneur, start by familiarizing yourself with generative AI tools. Here are some low-risk ways to incorporate AI into your editorial operations:

Experiment with ChatGPT or Claude:

• Use them for brainstorming grant proposals
• Simplify complex topics for your audience
• Explore their potential in various editorial tasks like copy editing, tone, style, etc.

Try YESEO: 

• This free Slack-based tool helps with headlines and SEO metadata suggestions. Created by RJI fellow Ryan Restivo, it’s used in over 400 local newsrooms and supports 10+ languages.

The more you experiment with these tools, the better you’ll understand their capabilities and limitations. This hands-on experience will help you develop best practices for AI use in your newsroom and identify potential pitfalls and how to avoid them. Start small, learn continuously, and scale up as you become more comfortable with the technology.

What’s the AI tool or workflow you’ve seen that has had the biggest impact on labor or time-saving?

Nikita: ChatGPT, when used to its potential, is one of the most effective tools available. Building Custom GPTs is an excellent way to leverage its capabilities for your newsroom, creating customized bots to handle specific tasks and boost productivity. Joe Amditis has an incredible guide on getting started with building Custom GPTs, which is a great read to better understand its potential.

What are the pluses and minuses of the big players (ChatGPT, Gemini, Perplexity, etc.) for local publishers? Is it a matter of personal preference, or is there a real working difference between these generative tools?

Nikita: There are major differences in the capabilities of each tool, and I use all of them, with ChatGPT, Claude, and Perplexity being tools I use multiple times a day. Here’s what I use each one for: 

• ChatGPT is one of the most advanced models and is great for reasoning tasks. You can create your own mini bots called CustomGPTs that can be instructed to do certain tasks, like helping write headlines, grant proposals, or sales emails.
• Claude has the best model out there as of its latest release, and I find it to be really good for writing tasks.
• Gemini: I use it mainly to chat with YouTube videos and documents in my Google Drive or Gmail.
• Perplexity has replaced Google for me since last year. This is where I go when I need factual information, as all the generated text has sources, and I can fact-check all the information.

How can I protect my data/my audience’s data/my IP and still make use of AI tools?

Nikita: Carefully read the terms and conditions of all the AI tools you are using and ensure that they are not using your data to train their models. If you are using ChatGPT, switch to their ChatGPT for Teams subscription, which provides you with enterprise-level security, and they don’t train on your conversations with ChatGPT. If you are on the free or even paid personal version of ChatGPT, they are training on all your conversations (unless you opt out of them).

---

This discussion with Nikita Roy took place in the News Entrepreneur Community Slack on August 8, 2024. If you are a LION member or interested in connecting with others who are working in the local news ecosystem, you can request an invite to join here.