As people increasingly turn to large language models for political tasks, including voting guidance, the political neutrality of AI chatbots has emerged as a major policy concern. American AI chatbots are used globally, yet little is known about their behavior as tools for political decision-making and potential political bias in non-U.S. contexts.

To address this gap, researchers ran an experiment during the final week of Japan’s February 8, 2026, general election. The experiment reveals a striking pattern: when asked which party to support in the election, five major AI models from three companies overwhelmingly directed voter profiles with left-leaning policy positions toward the Japanese Communist Party (JCP). The reason, according to the researchers, has to do with the information environment AI systems can access.

These findings, published in a working paper titled Why Do AI Models Tell Left-Wing Voters to Support the Communist Party?, “suggest that AI voting advice may be shaped as much by the information-retrieval environment as by model training, with implications for governance frameworks that rely on U.S.-centric assumptions,” write the researchers, Andrew Hall, the Davies Family Professor of Political Economy at the Stanford Graduate School of Business and a senior fellow at the Hoover Institution, and Sho Miyazaki, a visiting researcher at Waseda Institute of Political Economy, an incoming Ph.D. student in public policy at Harvard University, and a former predoctoral research fellow at Stanford University. Miyazaki is also a core member of the Stanford Japan Barometer, a project of the Japan Program at Stanford’s Shorenstein Asia-Pacific Research Center.

Sign up for APARC newsletters to receive our scholars' research updates >

How AI Models Deliver Political Advice in Japan: A Systematic Experiment

To understand how AI models provide political recommendations in the Japanese context, Hall and Miyazaki created 36,300 synthetic voter profiles with varying gender, region, and stated political views on 12 policy issues spanning security (constitutional amendment, defense spending, espionage law), diplomacy and immigration (China relations, foreign workers, permanent residency), energy (nuclear power), economic (consumption tax, social insurance), and social domains (dual surnames, restrictions on corporate donations, Diet seat reduction).

They then queried five models from three AI companies (OpenAI, Google, and xAI) during Japan’s February 8, 2026, Lower House election, asking each model to recommend a political party based on the voter profiles. All five models were queried with web search enabled and could access current information.

The researchers found that policy positions overwhelmingly dominated the models' party recommendations, producing swings of 50 to 98 percentage points in party choice, compared to just 0.5 to 7 percentage points for demographic factors. Thus, demographic effects are an order of magnitude smaller than policy effects.

Furthermore, left-leaning policy views in voter profiles caused all five AI models to converge overwhelmingly on recommending the Japan Communist Party, even though other parties hold broadly similar positions on the issues tested. The concentration on recommending JCP under left-leaning policy stances is therefore not explained by ideological distinctiveness.

In the control condition without policy input, models showed no uniform left-wing bias: three of the five models recommended the Liberal Democratic Party at high rates, and JCP shares were low for four of the five models.

“The key finding is that JCP recommendation rates rise sharply when policy positions are provided, which is the typical scenario when voters use these tools in practice,” write Hall and Miyazaki.

Information Environment Asymmetry

Why the JCP? The researchers traced the pattern to the sources AI models cite when making recommendations.

The JCP operates Akahata, a self-described daily newspaper published on a fully open website that AI web-search tools can freely access. In contrast, Japan's major news outlets have implemented technical barriers (known as robots.txt restrictions) that block AI crawlers from accessing their content, a move driven by copyright concerns.

The researchers found that the JCP's open website and party newspaper were among the most-cited sources in the AI models’ recommendations. Unable to distinguish between editorially independent journalism and partisan content, the models treated the JCP content as a credible news source. Thus, the information environment available to AI is systematically skewed toward the JCP's partisan sources that are designed to persuade rather than to scrutinize and inform.

“A model that retrieves information from jcp.or.jp/akahata and simultaneously classifies that site as news media is not simply making a labeling error: it is operating in an information environment where the boundary between party communication and journalism is genuinely blurred, and where the consequences of that blurring flow directly into its recommendations,” Hall and Miyazaki write.

The researchers also found that incorporating X search amplified left-leaning recommendations in Japan, the opposite of expectations based on the U.S. discourse environment.

Implications for Democratic Systems in the AI Age

The study's findings carry significant implications:

• AI governance frameworks should treat content access policy and AI political neutrality as deeply intertwined domains.
• Election commissions should create nonpartisan platforms that compile structured data about party positions so that the information is comparable, party-independent, and machine-readable.
• News organizations should recognize that by imposing copyright-motivated content access restrictions, they may inadvertently cede influence over AI-mediated information to partisan actors. They may wish to consider forms of negotiated access.
• Political actors will likely begin to optimize their communication for AI.
• Users should exercise caution in using AI as a voting advisor and be conscious of its potential biases and blind spots.

“If AI systems are going to act as political intermediaries more broadly, two problems need to be addressed,” writes Hall in an article about the research via his Substack. “The first is informational: ensuring that what the sources models read reflects the same balance of scrutiny and debate that voters encounter in a healthy media ecosystem. The second is advisory: deciding how an AI system should even translate a voter’s values into political guidance in the first place.”

Learn more about the Stanford Japan Barometer and its work  >

Introduction and Contribution:

Most people move for economic or personal reasons, such as attending college, starting a new job, or being closer to family. Accordingly, residential moves are often beneficial for movers: they can improve life satisfaction, offer movers new economic opportunities, and increase long-term earnings. However, what is often little acknowledged is that moving can generate significant political costs: movers may have to learn about the political issues salient in their new place of residence while simultaneously facing the daunting task of settling into a new place of residence. It is because of these challenges that moving can change the extent to which someone engages in politics.

In “The local costs of moving,” Hans Lueders proposes that to understand how moving affects political engagement, we need to distinguish between national and local engagement. Studying political engagement among movers in Germany, he finds that German movers remain similarly engaged in national politics but become considerably less engaged in local politics. Lueders argues that national engagement is unlikely to change much after a move because the political context remains the same. Intuitively, the same political candidates run for national office no matter where one lives, and the country’s most pressing political issues remain the same as well. By contrast, the political context changes significantly when it comes to local engagement: living in a new place means that movers have less political knowledge (e.g., of local political candidates or salient issues), limited social networks to facilitate local engagement, and a weaker sense of civic duty to engage. Lueders finds no evidence that movers adopt new norms or political ideas — mainly because Germans tend to move to places that are socially and politically similar.

Lueders draws our attention to how moving — and the disengagement it generates — can undermine local democratic accountability. Indeed, when movers cannot communicate their preferences to local leaders, what follows is “representational inequality” between movers and “stayers.” That domestic migration can add or remove 5-10% of a county’s population over a decade, thus has serious consequences for the quality of democracy.

Importantly, Lueders broadens the geographic scope of research on political engagement. Social science research on moving has been heavily informed by data from the US, where “strict voter registration requirements…have been described as more costly than the act of voting itself.” Indeed, the US’s unique — and uniquely burdensome — voting regime has been shown to weaken both local and national engagement for movers. Lueders’s research suggests that these findings do not travel beyond the US. In Germany and much of the Western democratic world, movers are legally required to register their new address with local authorities, and are then automatically added to the electoral rolls. This removes a key barrier to political engagement, at least at the national level. American readers may rightfully ask which interests are advanced or undermined by the current status quo.

Engagement Before and After Moving:

Lueders introduces two competing accounts of how moving affects political engagement. On the first account, moving imposes serious epistemological and social costs: movers must learn new information about politics, form new social networks, and come to see themselves as members of a new community. Not only does all of this take time, but movers usually prioritize more urgent personal matters — e.g., finding housing or childcare — such that politics takes a back seat. 

Weakened social ties mean that movers interact less often with people who could inform them about local issues, candidates, or initiatives — which are hard enough for longtime residents to grasp. Members of social networks also enforce norms of participation on each other; movers who lack social ties will thus be more content to abstain from voting or volunteering for campaigns. It could be inferred from this account that the further away one moves, the less engaged one will be with one's new home: candidates and issues seem even more novel, while social networks become even more fractured.

A second account highlights how the context of a new place can change engagement, as movers are exposed to new political ideas or norms around participation. This may be because movers are persuaded to approach politics differently, or for more instrumental reasons (e.g., if one’s preferred party already wins by large margins in the new place, engagement will seem less pressing). The contextual account assumes that moving entails a big change in one’s political environment.

Methods and Findings:

Lueders uses German household panel data collected between 1984 and 2020, which totals over 500,000 “respondent-year observations.” By comparing how engagement varies over time between movers and stayers, he can home in on the changes in engagement caused by moving itself, accounting for any baseline differences caused by the kinds of people who choose to move or stay. National engagement is measured by self-reported levels of national political interest, whether respondents voted in the last national election, and whether they plan to vote in the upcoming one. Local engagement is measured by self-reported attachments to one’s place of residence, how frequently they participate in local political and citizen initiatives, and their frequency of volunteering in local associations and organizations. 

Lueders’ findings are consistent with the first account, in which local engagement declines due to lower-quality information and weaker social ties. He finds no evidence that Germans’ levels of national engagement change, regardless of the distance of one’s move.

Image

Figure 2. Changes in engagement around moves of varying distances. This figure explores whether movers’ engagement in national (top panel) and local engagement (bottom panel) changes around moves of varying distance. Each coefficient reflects the estimated change in engagement among movers compared to the baseline (all stayers plus movers six or more years before a move). Vertical bars are 95% confidence intervals. Data from the German Socio-Economic Panel (Goebel et al., 2019).
 

By contrast, the V-shaped patterns in the lower panel show that local engagement changes significantly, declining in the lead-up (around five years) before a move, reaching its lowest point in the year of a move, and then slowly returning to pre-move levels in subsequent years, but without fully recovering. Importantly, engagement declines with distance, as the most local moves (i.e., within the same town or county) leave engagement largely unchanged.
 

Image

Figure 1. Changes in engagement before and after a move. This figure explores whether movers’ engagement in national (top panel) and local engagement (bottom panel) changes around a move. Each coefficient reflects the extent to which movers depart from the overall trend in engagement in a particular year before or after their move. Vertical bars are 95% confidence intervals. Data from the German Socio-Economic Panel (Goebel et al., 2019).
 

Against the “contextual” account, Lueders finds that the majority of German moves occur over short distances, which makes it unlikely that contexts differ dramatically. And indeed, most of the places to which Germans move are sociopolitically similar (to where they left) in terms of levels of turnout and federal election outcomes.
 

Image

Figure 3. Most moves occur over short distances. This figure uses data on all cross-county moves in Germany in 2015 to compute various metrics of the distance of such moves. Left panel: distribution of the distance between origin and destination counties. Center panel: share of all moves from a particular county that go to neighboring counties. Right panel: share of all moves from a particular county that lead movers to other counties in the same state. Own calculations using data from FDZ der Statistischen Amter des Bundes und der Lander (2019). The vertical dashed line indicates the median move (left) or median county (center and right).

Image

Figure 4. Movers tend to move between politically similar environments. This figure reports the distribution of the change in environments that movers experience upon a move (dark blue). This distribution is contrasted with the distribution of change in environments one would expect when simply considering population totals between county pairs (grey). The vertical lines indicate the medians for the actual (dashed line) and benchmark (dotted line) distributions, respectively.
 

Ultimately, “The local costs of moving” underscores how highly individual life events can undermine the quality of collective governance.

*Brief prepared by Adam Fefer