How we’re using topic modeling to understand news coverage of data centers
Data centers have become a growing part of news coverage across the country, bringing together stories about AI, energy, jobs, costs and local development. But with thousands of articles published across national and local outlets, it can be difficult to see what the coverage looks like as a whole.
That’s where the Automated News Coverage Tracker comes in. Developed at Northeastern University, the project uses computational methods to analyze large collections of news stories and explore how coverage differs across topics and regions.
For its first use case, the team is looking at AI data centers. We collected coverage from national news outlets and from Virginia, Ohio and Texas, using topic modeling to see what other subjects appear in articles about data centers and whether those patterns change depending on where the stories are published.
Looking at data center coverage
For the initial analysis, we collected articles mentioning “data center” from January through May 2026 using Media Cloud, an open source platform for researching media coverage. We started by comparing national coverage with stories from Virginia, Ohio and Texas.
Since every article in the dataset already mentions data centers, we were interested in what else appeared in the coverage. We used Latent Dirichlet Allocation, or LDA, to find groups of words that tend to appear together across the articles. This allowed us to look at topics such as energy and power, jobs, costs and taxes, local development and community concerns across the full collection.
But the computational analysis is only one part of the project. We also read articles from each group to see what those topics look like in the actual reporting and to check whether the patterns from the model make sense.
What we’re starting to see
to take a broader view, connecting data centers to major companies, investment, AI growth and the growing demand for energy. Local coverage brings the same story closer to individual projects and the issues surrounding them.
The patterns also vary from state to state. In Virginia, costs and electricity are a major part of the coverage, including questions about who should pay for the infrastructure needed to support data centers. In Texas, the coverage connects data center development to electricity and water use, tax incentives and local control. In Ohio, infrastructure and development feature prominently, including new power generation, zoning decisions and project approvals.
These differences show how the same data center boom can become a different story depending on where it is covered. At the national level, the story is often about the scale of AI growth and the energy and investment needed to support it. At the local level, those same developments become questions about costs, resources and what new projects mean for individual communities.
Where we go from here
The early analysis has given us several questions to follow, from differences between national and local coverage to who gets to speak in stories about data centers. The next step is to see which of these patterns continue to show up as we expand the analysis.
The larger goal of the Automated News Coverage Tracker is to create a workflow that can be reused beyond this one story. Data centers are our first test case, but the same approach could eventually be applied to other issues that generate large amounts of coverage across different places and news organizations.
The project has been selected for presentation at the Computation + Journalism conference in October, where the team will share the Tracker and what we have learned from using data center coverage as its first case study.
- What’s the real story behind the spread of AI data centers? It depends where you get your news. - September 22, 2026
- How we’re using topic modeling to understand news coverage of data centers - September 22, 2026
- Where Should Data Journalists Be This Spring and Summer? - May 6, 2026
