By 2023, an average of 2.5 local newspapers will close every week. More than half of U.S. counties now have little or no reliable local news coverage, and that trend is accelerating.
This is a business problem. It is definitely also a question of democracy. For centuries, local journalism has engaged voters in local politics and held politicians accountable to those voters. Small newspapers that conducted persistent research also regularly reported on stories of national importance. patriot news For example, uncovering the Jerry Sandusky scandal at Penn State.
What is the answer to this crisis? “Everyone says, ‘Let’s use AI to help,'” says Monica Lamb, a computer science professor at Stanford University. The problem with this, she added, is that most AI tools are unreliable. She cited a 2025 study in which the BBC used a leading AI model to analyze news content on its website. According to the BBC, more than half of the responses from the AI had “serious issues” such as factual errors or fabricated quotes.
“It’s not that easy,” Lamb said.
Lam is currently working with technologists and journalists to develop tools that are more useful for the news industry. In collaboration with Cheryl Phillips, founder of Stanford University’s Big Local News, and with seed funding from the Stanford Human-Centered AI Institute and grants from Stanford and Columbia University’s Brown Institute for Media Innovation, Lamb created DataTalk, a chatbot specifically designed to help investigative journalists and cash-strapped news workers do their jobs more efficiently without sacrificing factual accuracy. DataTalk is built on a large language model and is designed to capture and analyze information stored in large and sometimes unwieldy public databases.
“Journalism is losing a lot of people, and thorough investigative work is harder than ever,” Lamb said. “Our hope is that as more people learn about the tools we are building, and that we can continue to improve them and generate success stories, our hope is to strengthen this type of journalism for years to come.”
What is Data Talk?
Investigative journalists often rely on knowledge of database languages like SQL and the expertise of data scientists to uncover important stories. With DataTalk, you can simply type your question into the chat window and get an answer within seconds.
The tool, which is available to everyone, is currently focused on campaign finance data. That means public access is limited to questions related to federal political campaigns, such as how much money a candidate for Congress has raised from outside the state.
But the tools are expanding. of baltimore banner recently began using DataTalk to uncover news stories buried in 311 non-emergency call log data. In the coming months, Big Local News hopes to work with Lam and other journalism organizations to identify other key datasets that can be integrated into DataTalk and create a system that will make it easy for local journalists to add their data to the agent. State-level campaign finance records are one example.
Along with the analysis, DataTalk provides the code used to perform the analysis and a plain English explanation of what the code is doing. This ensures that what you are asking in technical jargon is the same question the journalist would have asked in plain language. We also discuss ways in which your analysis may be limited.
To ensure DataTalk’s accuracy and usefulness, Lamb and Phillips worked with subject matter expert Derek Willis, one of the nation’s leading campaign finance data journalists, to help refine how the chatbot performs search and interpretation.
“Willis was able to provide critical instructions to make sure that when a lay journalist asked an agent a question, the agent knew which table to go to and how to formulate a query from the general instructions they received,” says Phillips. “Simple data sets like 311 calls may not require this level of expertise. We consider the structure of the information and the area we are looking at to determine what expertise is needed to ensure this model works.”
Once the tool was established, Lam’s group worked with Willis and the Big Local team to continually evaluate and improve the DataTalk interface. He also worked with students in Phillips’ class to deepen their understanding of how the agent works, and has continued to maintain and improve the tool’s technical infrastructure ever since.
From the classroom to the newsroom
In the fall of 2024, Phillips piloted a chatbot in her “Big Local Journalism” class. The students partnered with local newsrooms to publish three stories throughout the quarter, focusing on stories about campaign finance. One article compared the donor pools for two candidates in Hawaii’s state legislative race. Another article highlighted Kamala Harris’ campaign spending on reproductive health ads in Georgia. (Students manually fact-checked each story and used their own code to reproduce DataTalk’s analysis.)
“The news outlets that published this piece were happy to do so,” Phillips said. “These were stories that wouldn’t have been told any other way.”
Around this time, main monitor We reached out to do our own analysis comparing campaign contributions from within and outside the state. Reassured by the success of the pilot, Phillips helped reporters with interviews. monitor Do her research.
AI toolbox for journalism
DataTalk is part of Phillips and Lamb’s broader plan to support the world of investigative journalism. What they have in mind is a complete toolkit of applications to help newsrooms produce stories, whether they are small local operations crumbling under the strain of scarce resources or national newsrooms with deep investigative power. The academics will also provide tutorials on how to use these applications, as well as different stories that may be assigned to them.
Next, the team would like to add DataTalk functionality to Agenda Watch. This feature uses computational methods and AI to collect meeting agendas and minutes from city councils, school boards, and other local decision-making bodies across the United States. Agenda Watch can also alert users to newsworthy items appearing in local documents.
“Taken together, this initiative aims to reduce the cost of producing responsible journalism,” Phillips said. “We hope this will allow us to dig deeper into the investigation and produce important stories.”
