The “techlash” against artificial intelligence is spreading, driven by workers’ fears of the technology’s potential to disrupt their jobs and upend their livelihoods.
While such fears appear to be widespread, a closer understanding of where AI-exposed workers are concentrated and what political party they align with can be illuminating, as AI policy is poised to play an important role in this November’s elections. Yet research on AI’s “political geography” is narrow. Broad surveys suggest mixed or unclear disparities in U.S. voters’ views about the technology.
Just a few weeks ago, a survey from the University of Pennsylvania’s Annenberg Public Policy Center flagged new pessimism about AI, a bipartisan demand for regulation, and a stronger tilt among Democrats toward intervention. In February, Data for Progress found opinions closely divided, with 48% of likely voters viewing AI favorably and 46% viewing it unfavorably. Around the same time, academic researchers Nicholas Bloom and Christos Makridis found that Democrats are more likely to use AI and to be employed in jobs with higher AI exposure. And last year, scholars Alexis Antoniades, Carlos Felipe Balcazar, Emmanouil Chatzikonstantinou, and Andreas Kern showed that counties with more job postings containing AI-related skills tend to have a higher Democratic vote share.
This report adds to this limited understanding by analyzing the political behavior of voters in the locations where jobs are becoming involved with AI, as indicated by data from Anthropic. Such a look at how local jobs’ AI exposure lines up with local voting patterns in the last presidential election may offer signals about where and for whom AI could become an especially hot political issue.
Local voting patterns in the last presidential election offer signals about where and for whom AI could become an especially hot political issue.
So, what does the AI political map look like, with its implications for sentiments toward AI and subsequent voting behavior?
To explore the possible interplay of AI diffusion and political partisanship, we build here on Brookings’ earlier research to leverage AI “exposure” statistics for occupations as a way to map patterns of local AI impact against county election outcomes in the 2024 presidential election.
In the earlier report, we used data from OpenAI to show that AI excels at supporting or executing activities such as conducting research, writing code, preparing analyses, creating marketing content, and drafting presentations—the types of highly cognitive, nonroutine tasks that better-educated, better-paid office workers carry out. Given that, the report found that the more involved workers are in high-level office or information-based work, the more involved they will be with AI. All of this has implications for the geography of AI engagement, with big city areas containing many of the white collar office workers being most affected.
In this analysis, we draw on similar AI exposure estimates from Anthropic based on actual usage of their Claude model, which weighs automative uses (in which the model completes the task with little user input) twice as heavily as augmentative uses (in which humans collaborate with AI through learning, iterating, and validating to understand and get tasks done). We then construct a county-level exposure measure by weighting these occupation-level estimates by local employment numbers, so that each county’s exposure estimate reflects how its particular mix of occupations is involved with AI.
Along these lines, we again find that the more involved workers are in office or information-based work, such as computer programming and marketing, the more involved they will be with AI. This again reinforces the urban focus of AI activity, which in turn gives hints toward the political nature of the places most exposed to AI.
So, what do we find about the political geography of AI? First, we would caution that the politics suggested in our analysis remain an implied byproduct of “where people work and what skills they hold,” as Bloom and Makridis write, rather than a specific fact of actual ideological perspectives.
With that said, our analysis reveals a strong relationship between AI involvement and local political partisanship. Specifically, our analysis reveals a strong correlation—albeit one without causality—between a county’s AI automation exposure score and its Democratic vote in the 2024 election.
Figure 1 shows that 62 of the 100 most AI-exposed counties in the nation went “blue” in the 2024 presidential election. These counties make up 75% of the population of those 100 most AI-exposed counties, and between 14% and 19% of workers there are in occupations where AI is both theoretically capable of handling tasks and already being used to automate work more than augment it.
Given those patterns (and acknowledging the lack of directly measured partisanship), highly exposed blue counties such as New York County; Broomfield County in the Denver area; San Francisco County; Boulder County, Colo.; Santa Clara County, Calif.; Hennepin County in the Minneapolis area; and King County in the Seattle area—and places like them—could be locations where AI turns out to be an object of special anxiety.
The association between AI exposure and voting behavior is similar at the state level. States with concentrations of AI-exposed work, such as Massachusetts, New York, California, and the Washington, D.C. area—which are dominated by big, AI-oriented metro areas—exhibit observed AI exposure levels ranging from 13% to 17%, and tend to have voted blue in 2024.
At the same time, states that lack large, AI-involved metro areas display much lower exposure scores and tended to vote “red” in 2024. These include Nevada, Mississippi, Wyoming, and North Dakota, which have AI exposure levels between 10% and 11%.
The last presidential election’s swing states provide an interesting perspective as well. Most notably, Arizona and Georgia display very high AI exposure levels (and are among the 15 states most exposed to AI). For their part, Michigan, North Carolina, Pennsylvania, and Wisconsin exhibit moderate AI involvement. Nevada, in contrast, displays very low AI exposure, ranking last among states.
As to what all of this means, it’s important to note that this analysis does not imply that blue counties and states are facing immediate job dislocation from AI automation, or that anxiety about AI automation in these places is by itself beginning to make more people vote Democratic. (Indeed, Antoniades, Balcazar, Chatzikonstantinou, and Kern present evidence that AI adoption generates jobs, attracts educated workers, and actually benefits Democrats.) What’s more, many longer-standing factors such as education levels in these places clearly prompt Democratic voting.
Yet with that said, the simple correlation of AI automation and Democratic voting does suggest—as the economist Jed Kolko observed about industrial automation a decade ago—that AI exposure may turn out to be a source of economic and social concern especially in blue counties and states going forward.
Simply put, on average, blue places concentrate workers in numerous occupations in which workers are right to feel more anxiety about AI-driven job dislocation than workers in red places. Therefore, going forward into this November’s midterm elections and beyond, America’s bluest counties may become hotbeds of some of the AI era’s most agitated voters.
Authors
The Brookings Institution is committed to quality, independence, and impact.
We are supported by a diverse array of funders. In line with our values and policies, each Brookings publication represents the sole views of its author(s).
