Why we built an audience-focused research project to test AI chatbots for local news 

At the Center for Innovation and Sustainability in Local Media, one question drives much of our work: what makes local journalism work — and how can we help it work better? The answers change over time, but the mission stays the same: connect journalists, audiences and organizations in ways that sustain both civic life and local news itself. 

Lately, that means looking closely at audience demand. Local news has a demand problem, with audiences seeking information in a different way than is often provided. Like any business venture, an organization needs to meet its consumers where they are. 

We wondered: How can AI influence demand? Are news consumers looking for a way to engage more interactively? Additionally, are AI-powered chatbots an easy way to tackle persistent problems faced within newsrooms?  

We believe in research grounded in real-world practices and newsroom realities. That’s why we launched the Local NewsBot Studio, a hands-on pilot program in which we partner with newsrooms to design and build custom chatbots. Our goal is not just to create working tools, but to study how audiences use them and how newsrooms can adapt AI to meet local needs. 

What the Local Newsbot Studio is 

The Local NewsBot Studio brings together research, product development and newsroom collaboration. Over several months, our team at CISLM has worked side-by-side with four news organizations across the Southeast — spanning community radio, digital-only sites and a legacy print newspaper — to co-develop chatbots tailored to their unique audiences. 

Each chatbot is built on the newsroom’s own reporting and resources, not general web scraping or open-ended AI. That means every bot is trained on trusted, curated information specific to its community, using only the information provided by the news organization. 

Here’s what that looks like in practice: 

• Atlanta Civic Circle launched Ask ACC, a bot designed to answer civic and political questions leading into the 2025 local elections. The nonpartisan newsroom focuses on informing and inspiring metro Atlanta residents about civic life, and the bot helps extend that mission. 

• Chapelboro introduced Chappy, a friendly Clippy-like assistant that helps readers navigate local coverage, summarize stories from the past year and answer FAQs about the organization itself. 

• Henrico Citizen, out of Henrico County, Virginia, developed Henry, another site guide designed in the spirit of Clippy. Henry answered reader questions about the site, synthesized county-specific stories and made local coverage easier to access. It also responded to customer service questions such as “how do I subscribe?” 

• The News Reporter in Whiteville, N.C., built the News Reporter Help Desk bot. For the Pulitzer Prize-winning newspaper with a small but mighty staff, the bot took on “front office” duties — answering questions about subscriptions, placing ads or obituaries and other reader services. 

Local News Researcher Yanan Sun and I built these bots from start to finish — mapping audience needs to technical design, facilitating content curation for knowledge bases, and supporting newsroom staff on how to embed and maintain their bots. 

This combination of hands-on support and research insight is what makes the Studio different. It’s not experimentation for tech’s sake, but a deliberate experiment in service of journalism. 

Why chatbots? 

We’re building on what we learned from our own AI-powered chatbot, the Local News Audience Assistant, with a unique team with backgrounds in product and research that makes this possible: 

• Local News Researcher Yanan Sun is a former entrepreneurial fellow at the Brown Institute for Media Innovation and has co-founded an AI product that helps users read mobile apps’ privacy policies. 

• I built our Local News Audience Assistant chatbot for CISLM, which compiles case studies, best practices, Q&As, tip sheets, audio and video recordings and more around audience development from journalism support organizations. 

• Former Interim Director Jessica Mahone is a celebrated researcher in local news. 

Our team saw chatbots as the best way to experiment because they meet several pressing needs for local newsrooms: 

• Serving audiences directly. Chatbots offer a simple, intuitive way for people to find the information they care about. Whether that’s local election details, obit policies or guidance on how to subscribe, bots can provide answers quickly and clearly. 

• Surfacing valuable reporting. Many news organizations hold decades of community information in their archives. Chatbots can make those archives discoverable again. 

• Saving newsroom time. With small staffs often stretched thin, chatbots can take on repetitive customer service questions, freeing journalists to focus on reporting. 

• Creating feedback loops. The questions audiences ask chatbots provide valuable signals about what people are seeking from local news — insights that can inform editorial and business strategy. 

But our work isn’t only about proving chatbots “work.” We’re equally invested in understanding how audiences engage with them, what kinds of trust barriers might emerge, and what local journalists can learn from the data. 

Keeping it human — and local  

AI comes with well-documented risks. At CISLM, we’re committed to building and testing these tools responsibly, always with trust and transparency in mind. 

That’s why we designed the Studio around three principles: 

• Grounding in trusted content: These chatbots are powered by hand-curated spreadsheets and newsroom-provided material. 

• Human oversight: Both our team and the newsroom partners actively monitor bot performance, step in to correct errors and refine responses as needed. 

• Sustainability: Each newsroom leaves the program with a chatbot they can continue to use if they choose, plus analytics on how their audiences engaged with the product. 

By keeping the scope human-led and newsroom- specific, we’re ensuring these tools feel like extensions of local journalism rather than replacements for it. 

What we learned  

The Local NewsBot Studio wasn’t about proving whether AI could “solve” local news. It was about testing, in real newsroom contexts, where chatbots add value and where they fall short. In 45 days across four bots, we saw both promise and clear limits. 

• Small but possible. Each bot was designed and deployed in less than a month, with no in-house technologists required. For resource- strapped newsrooms, that’s a powerful signal. Experimenting with AI doesn’t have to mean heavy infrastructure or major costs. 

• Narrow works better. Bots with a clear purpose — like handling FAQs, customer service questions or helping audiences navigate archives — showed the strongest results. “Ask me anything” bots, in contrast, were less effective and more likely to frustrate users. 

• Engagement was modest but telling. Over the pilot, the bots logged 185 queries. While that’s not a tidal wave of interaction, it gave us valuable insight into what audiences expect. Content-focused bots generated more follow-up questions, suggesting that when users did engage, they wanted to keep going. 

• Gaps matter. About a third of conversations ended with an “I don’t know.” Questions that required real-time updates, like “What’s the latest story?” often tripped up bots without auto-updating knowledge bases. Each miss chipped away at trust. 

• Trust is fragile. Some users assumed the bots were powered by general AI, not curated newsroom content. When the bots answered incorrectly or failed to return timely information, it wasn’t just a technical hiccup — it risked damaging the credibility of the outlet itself. 

• Mixed newsroom impact. Some partners found real value — one newsroom reduced repetitive customer service calls, another provided civic information in a high-stakes election year. Others reported limited utility, especially when accuracy or site integration fell short. 

• Maintenance is the barrier. Without automated links to publishing systems, bots quickly became outdated. For newsrooms with lean staffs, the ongoing upkeep proved to be the hardest part of the experiment. 

Looking ahead 

We found that the long-term viability of newsroom chatbots will depend less on whether they “work” in the abstract and more on the strategy behind them: clearly defined use cases, auto-updating content pipelines, transparency about limitations and systems for gathering structured user feedback.  

The takeaway? Chatbots can be affordable, practical tools for local news — but only when built with a focused scope, clear strategy and sustainable plan for maintenance.