What happens when local news audiences interact with AI-powered chatbots on their favorite newsroom websites?
That’s the question we set out to explore through the Local NewsBot Studio, a first-of-its-kind experimental project from CISLM. Working with four local newsrooms across the Southeast, we helped build and test chatbots designed to serve both the newsroom and its readers. Here’s what we learned about what audiences want, how they engage, and what this means for the future of local journalism.
1. Most readers want quick answers — then they leave
Across 185 chatbot sessions over 45 days, a clear pattern emerged: more than half of users asked one question, received an answer, and left. That may seem like a short interaction—but for audiences looking for a quick, useful response, it might be exactly what they needed.
More than half of the conversations were a single question-and-answer exchange… which suggests users got what they needed and moved on.
2. Content-focused bots sparked more curiosity
Not all bots were the same. Chatbots built to deliver journalistic content (like Chapelboro’s or Atlanta Civic Circle’s) saw much longer conversations than bots focused on customer service (like The News Reporter’s). For readers interested in politics, community events, or recent reporting, these bots became a mini search engine.
43% of content-based chatbot sessions had follow-up questions, compared to almost none for customer-service bots.
3. Audiences asked for more than news
Some audience questions aligned perfectly with newsroom beats—like city council updates, election info, or community events. But others got more creative: people asked about cooking, poetry, math, and even the chatbot itself (like “Who is Henry?” at Henrico Citizen).
Users also diverged from news altogether, asking about math, poetry, cooking, and more.
4. Users do click on chatbot prompts — if they’re relevant
About a third of users clicked on a prompted question (pre-written by the newsroom) instead of typing their own. But this varied: Chapelboro users mostly asked their own questions, while Henrico Citizen readers leaned heavily on prompts—especially when they were clearly useful or entertaining.
Prompt questions shaped engagement, and relevance really mattered.
5. Readers value accuracy—and can tolerate “I don’t know”
We deliberately limited each chatbot’s knowledge to content from the newsroom itself. That meant some user questions, like “Who is the President?” didn’t get answers if that topic hadn’t been covered. Still, newsrooms and audiences alike preferred a chatbot that says “I don’t know” over one that makes things up.
About 1 in 3 sessions included an unanswered question, but accuracy was the top priority.
These bots weren’t perfect, but they offered a glimpse into what audiences are looking for from local news organizations. They want fast, accurate answers. They’re curious about their communities. And they’re open to engaging with AI, if it’s helpful, honest, and relevant.
Explore more audience insights in the full Local NewsBot Studio Report.