"Local NewsBot Studio Report" table of contents
Local NewsBot Studio Report: How We Built Chatbots with Four Local Newsrooms
Yanan Sun, CISLM Local News Researcher
Why we built this
As artificial intelligence becomes more accessible and widely adopted across industries, it is also reshaping journalism in workflow automation, content production, product innovation and more. News publishers have invested in generative AI, including custom-built chatbots, to regain lost traffic to zero-click search results from apps like ChatGPT and Gemini. Organizations such as the Washington Post, Gannett/USA Today, and TIME Magazine have all developed generative AI tools trained on their own published reporting to maintain reader engagement.
Beyond audience-facing products, AI has been adopted in newsroom operations behind the scenes. A study on news articles across the globe found a significant increase in newsrooms’ use of AI in writing automation (73% of news organizations), data analysis (68%), and personalization (62%). Researchers from MIT built an AI-powered audience feedback analysis tool to interpret comments from multiple channels, generating insights synthesizing audience comments.
Institutional and financial support has played a key role in advancing AI innovation in local newsrooms. Initiatives led by organizations like American Journalism Project, the Lenfest Institute, and JournalismAI have provided funding and resources for local newsrooms to experiment with AI projects. However, smaller local newsrooms often lack access to these opportunities. Many are interested in AI but face barriers including limited time, financial resources, staffing, and technical capacities.
In 2024, the Center for Innovation and Sustainability (CISLM) built the Local News Audience Assistant, a chatbot that answers questions about best practices of audience engagement. It responds to queries on topics such as growing newsletter subscriptions, improving social media presence, developing event strategies, and more. Led by our Project Manager Sarah Vassello, the tool was built using manually curated resources about local news audience research from 2018 onward. It provided actionable guidance to local media outlets and shared related resources based on user questions.
Drawing on our experience in product development and technical implementation, we partnered with four small local newsrooms to collaboratively build AI-powered chatbots tailored to their needs, creating the Local NewsBot Studio. In addition to helping newsrooms experiment with AI, we used this program as a research opportunity to document local journalists’ goals, concerns, and reflections around AI adoption. We hope these insights can inform future efforts to responsibly integrate AI in small local newsrooms.
Key takeaways from the report:
- Feasible for small newsrooms: Chatbots were built and deployed in less than a month at a low cost, showing that AI tools can be affordable and accessible without in-house technologists.
- Best for focused use cases: Bots addressing FAQs, customer service, or archive navigation provided the clearest value; broad “ask anything” bots struggled more.
- Engagement was limited but instructive: 185 inquiries across four bots in 45 days; content-focused bots saw more follow-up questions than service bots.
- Knowledge gaps reduced effectiveness: About 32% of conversations triggered “I don’t know,” and bots often failed with time-sensitive queries like “latest stories.”
- Audience trust is fragile: Users sometimes expected general AI answers; incorrect or outdated responses risked credibility.
- Mixed newsroom experiences: Some saw strong civic or customer service value; others reported limited utility due to accuracy issues or site integration challenges.
- Maintenance is the biggest barrier: Without automated updates tied to publishing systems, bots quickly became outdated.
- Future success depends on strategy: Sustainable bots require auto-updating knowledge bases, clear scope, transparent limitations, and structured user feedback.
Public skepticism about AI in journalism also shaped our interest in studying this space. Research from the University of Minnesota found that more respondents had little or no confidence in newsrooms using AI than those who reported some or a great deal of confidence. A study by the Pew Research Center also showed that only 10% of U.S. adults said that the impact of AI on the news over the next 20 years would be somewhat or very positive.
Recognizing the skepticism, our pilot also aimed to understand how local news audiences interact with AI in a real newsroom setting. By analyzing user behavior and chatbot engagement data, we explored how readers perceive AI tools and whether such interactions influence their information-seeking habits.We hope this analysis helps researchers, journalists, and product designers understand news consumers’ interactions and supports trustworthy, cautious AI experiments in journalism.