Pew AI Survey Finds Democrats More Worried Than Republicans as Job Fears Rise
Pew Research Center has recorded a first: Democrats are now more concerned than Republicans about AI’s growing role in daily life. The new Pew AI survey puts Democratic concern at 56%, compared with 49% among Republicans.
The seven-point gap is modest, but the direction is historically significant. Republicans had expressed more concern in every survey from 2021 through 2024. The parties were effectively tied in 2025.
The sharper reversal concerns employment. Seventy-five percent of Democrats now expect artificial intelligence to produce fewer American jobs over the next 20 years. That compares with 68% of Republicans and only 58% of Democrats two years earlier.
These results do not prove that AI has become a decisive partisan issue. They show that its political meaning is changing. The debate is moving from abstract innovation toward jobs, inequality, energy use, and control over deployment.
That shift creates pressure for both parties. Democrats must translate concern into a credible policy without appearing hostile to useful technology. Republicans must defend rapid development while addressing voters who expect AI to eliminate jobs.
The Pew AI Survey Captures a Partisan Reversal
Democratic concern has climbed while Republican concern has declined, reversing a pattern that lasted through most of the generative AI boom.
Pew surveyed 3,488 American adults between June 22 and June 28, 2026. Participants belonged to its American Trends Panel, a probability-based panel designed to represent the adult population.
The Pew AI survey asked whether increasing AI use in daily life made people more concerned, more excited, or equally concerned and excited. Democrats and Republicans included independents who lean toward either party.
Across all adults, 52% said they felt more concerned than excited. Another 37% felt the two emotions equally, while only 9% felt more excited than concerned.
The partisan movement underneath that national result is the real story. Democratic concern rose from 31% in 2021 to 46% in 2023, then reached 56% in 2026. That represents a 25-point increase over five years.
Republican concern followed a different path. It stood at 45% in both 2021 and 2022, rose to 59% in 2023, and then declined. It reached 49% in the latest survey.
The change is not evenly distributed within either coalition. Concern among liberal Democrats rose from 45% in 2023 to 63% in 2026. Meanwhile, the share among conservative Republicans fell by 14 points over that period.
Pew found no significant comparable movement among moderate and conservative Democrats. Views also remained relatively stable among moderate and liberal Republicans.
That distinction matters because a partisan average can hide the source of political energy. The reversal is being driven mainly by each party’s ideological edge, not by simultaneous movement across every faction.
The result also differs from ordinary partisan polarization. On many issues, one party becomes more alarmed while the other becomes more dismissive because elected leaders send opposing signals. Here, concern rose sharply among Democrats after having already surged among Republicans in 2023.
Republicans did not become broadly enthusiastic about AI. Only 10% said they felt more excited than concerned, almost identical to the 9% recorded among Democrats. The difference concerns the balance between worry and mixed feelings.
Forty percent of Republicans described themselves as equally concerned and excited, compared with 34% of Democrats. The political divide therefore separates stronger concern from ambivalence, not rejection from enthusiasm.
That makes the Pew AI survey a measure of changing risk perception. It does not reveal whether respondents want regulation, slower development, worker protections, or limits on specific systems. Those policy questions remain open.
AI Job Loss Fears Are Driving the Larger Divide
The employment findings reveal a deeper shift than the seven-point gap in general concern.
Seventy-one percent of American adults expect AI to result in fewer jobs over the next 20 years. Only 5% expect more jobs, while 10% believe it will make little difference. The remainder were unsure or did not answer.
In 2024, 64% expected fewer jobs. The national measure has therefore risen seven points in two years, even as companies expanded workplace access to generative AI.
The partisan trajectory is more revealing. In 2024, 70% of Republicans expected AI-related job losses, compared with 58% of Democrats. Republicans held a 12-point lead in pessimism.
By 2026, the Democratic figure had risen 17 points to 75%. Republican expectations changed little, declining two points to 68%. A 12-point Republican lead became a seven-point Democratic lead.
The shift cannot be explained by Republicans suddenly expecting an employment boom. Only 7% now think AI will produce more jobs. That is slightly higher than the Democratic result of 4%, but optimism remains marginal in both parties.
These AI job loss fears connect an unfamiliar technology to a familiar economic question. Voters may not follow model benchmarks, inference costs, or AI safety research. They understand the possibility that software could reduce hiring, compress wages, or eliminate parts of a job.
Consider a routine knowledge-work scenario. A company can give managers software that summarizes meetings, drafts reports, reviews support tickets, and searches internal records. Each feature can save time without immediately eliminating a position.
The calculation changes when executives use those savings to freeze hiring or consolidate teams. Workers then experience AI as an employment decision, even when no machine fully replaces a person.
That distinction complicates public debate. AI can increase an individual employee’s output while reducing the number of employees an organization believes it needs. Productivity and job security are not interchangeable outcomes.
The concern also extends beyond direct replacement. Workers may fear weaker entry-level hiring, closer performance monitoring, lower bargaining power, or faster expectations. None requires a system capable of automating an entire profession.
Pew’s global employment study places the American finding within a wider pattern. In 34 of 37 surveyed countries, people were more likely to expect fewer jobs than more jobs.
The pessimism was especially pronounced in several wealthier countries. Around seven in ten adults or more expected job losses in the United States, Australia, and South Korea.
Economic inequality reinforces that anxiety. Pew found that Americans under 35 were more likely than adults over 50 to predict AI would widen the wealth gap. The respective shares were 56% and 38%.
Those perceptions do not forecast the labor market with scientific precision. They do reveal the standard against which companies and policymakers will be judged. Promises of productivity will receive greater scrutiny when most voters expect the gains to reduce employment.
Democrats’ AI Concerns Now Carry Political Pressure
The reversal pushes Democrats toward stronger safeguards, but public anxiety does not automatically produce a coherent governing program.
The survey arrived as AI moved closer to the center of the 2026 political debate. Candidates have faced questions about data centers, electricity costs, employment, safety, and federal oversight.
Democratic politicians now have a receptive audience for arguments about worker protection and corporate accountability. Yet the coalition’s concerns span several issues that require different policy responses.
Job displacement could lead lawmakers toward retraining, wage insurance, stronger unemployment systems, or rules governing workplace automation. Privacy concerns require limits on data collection and surveillance.
Data-center disputes involve electricity prices, water use, local permitting, and environmental effects. Frontier-model safety raises separate questions about testing, security, and federal authority.
Treating all those concerns as one generalized fear of AI would produce an unclear agenda. Voters may support restrictions on one deployment while continuing to use another every day.
Democratic leaders also face a credibility test. Warning that AI will remove jobs is easier than explaining which interventions can protect workers without blocking beneficial tools.
A blanket pause would encounter opposition from technology companies, researchers, business customers, and national-security officials. A purely voluntary framework could look inadequate to voters expecting substantial disruption.
Recent events show the tension. According to federal policy reporting, Democratic lawmakers have pressed for a stronger response while Republican leadership has resisted broad federal controls.
Some technology executives have also requested greater oversight, although their preferred rules do not necessarily match labor-focused proposals. Companies often emphasize extreme safety risks, national competition, or standardized evaluations.
Workers may prioritize something more immediate. They want to know whether an employer must disclose automation plans, how performance data will be used, and who receives the financial gains from productivity.
That gap between elite safety debates and workplace anxiety could determine whether Democrats convert concern into political advantage. A policy centered only on distant catastrophic risks may miss what the survey measures most clearly.
The party must also account for its own connection to the technology sector. Democratic administrations and lawmakers have supported research funding, domestic chip manufacturing, and technology-led economic growth.
That history makes a total anti-AI position unlikely. It also creates an opening for a labor-centered approach that distinguishes adoption from unmanaged displacement.
Republicans face a different pressure. A pro-development message aligns with competition against China and opposition to federal regulation. It becomes harder to sustain when Republican voters also expect widespread job losses.
Sixty-eight percent is not a fringe constituency. It is a large majority of Republicans anticipating fewer jobs, even if their general concern has declined since 2023.
The political contest is therefore not innovation against fear. It is a dispute over which risks deserve action, which institutions should respond, and whether rapid deployment can coexist with economic security.
Republicans Are Less Worried, Not Broadly Optimistic
Republican sentiment has moved toward greater tolerance of AI, but the data do not show mass confidence in its economic benefits.
The decline in Republican concern can support several interpretations. Conservative voters may be responding to party leaders who frame AI development as a national competitiveness issue.
They may also associate regulation with Democratic governance, technology restrictions, or slower competition against China. That framing can reduce support for intervention without increasing confidence in AI itself.
The survey cannot establish which explanation is correct. It measures changes in attitudes, not their causes. Pew’s data show the movement but do not isolate the messages, events, or personal experiences behind it.
The internal split matters. Conservative Republicans drove much of the decline, while moderate and liberal Republicans showed no significant change since 2023.
That pattern suggests political cues may influence the most ideologically engaged voters. It does not prove that leadership messaging caused the change.
Republican leaders also lack a simple position. Some prioritize fast development and oppose federal guardrails. Others have criticized specific companies, called for investigations, or supported restrictions on data centers.
Local politics can override national ideology. A voter may favor American leadership in AI while opposing a computing facility that affects local power rates, water supplies, or land use.
A separate voter survey illustrates that tension. It found 61% opposition to new AI data centers, including 75% of Democrats and 47% of Republicans.
The same poll found Republicans holding a narrow advantage on which party voters trusted to handle AI. The result was 42% for Republicans and 40% for Democrats, within the survey’s margin of error.
Those findings warn against reducing AI politics to a single left-right scale. A person can dislike data centers, fear job losses, oppose sweeping federal regulation, and still use AI tools.
The technology touches too many interests for one attitude to govern every decision. Consumers evaluate convenience and accuracy. Workers assess employment security. Communities consider infrastructure costs. Governments weigh safety and national competition.
Republicans can therefore become less concerned in a general question while remaining pessimistic about jobs. The emotional direction and the economic forecast are related, but they are not identical.
That distinction challenges claims that either party has secured durable ownership of the issue. Republicans have a narrow trust advantage in one poll, while Democrats show greater concern in another.
Neither finding guarantees an electoral benefit. Voters must first rank AI highly enough for it to influence their choices, then believe one party offers a practical response.
The New York Times and Siena survey found fewer than 1% of likely voters naming AI or data centers as their most important election issue. The economy remained far more important.
AI could still enter the election indirectly. Higher electricity bills, slower hiring, community development fights, or visible layoffs may be classified as economic issues rather than technology concerns.
That is why the employment result deserves more attention than a simple partisan score. The political impact will depend on whether voters connect their economic experiences to AI deployment.
What the Numbers Do Not Establish
The Pew AI survey identifies a real change in opinion, but it cannot explain every cause or predict specific policy choices.
First, the survey question asks about concern and excitement in daily life. It does not ask respondents to evaluate a particular model, company, law, or workplace system.
Two Democrats who select “more concerned” may want entirely different responses. One may support strict federal licensing. Another may want worker protections while welcoming ordinary consumer applications.
The same ambiguity applies to Republicans. Someone can oppose regulation because of national competition while remaining worried about employment. Another respondent may distrust technology companies but prefer state action.
Second, the seven-point partisan gap should not obscure the broader national direction. Majorities or large pluralities across political groups express concern, and very few are primarily excited.
This is not a story in which one party embraces AI and the other rejects it. It is a story about the intensity, framing, and recent direction of anxiety.
Third, the survey was conducted in late June but published in September. It predates some of the most intense September arguments about AI safety, midterm politics, and federal action.
Later news could move public opinion again. The timing does not undermine Pew’s result, but it limits claims about how voters reacted to subsequent developments.
Fourth, predictions about employment are perceptions rather than measured job losses. Respondents were asked what AI would do over 20 years, a period long enough to include economic cycles and policy changes.
The answers can shape politics even if the forecasts prove inaccurate. Employers, candidates, and AI companies must respond to perceived risk before long-term labor statistics settle the question.
Fifth, party averages can conceal demographic differences. Age, education, awareness, income, and ideology all shape how people evaluate AI.
Pew’s international work found that higher awareness sometimes corresponds with greater optimism about AI. It also found that higher awareness can coincide with stronger expectations of job losses in some countries.
Those results are not necessarily contradictory. People who know more about AI can appreciate its usefulness while taking its labor effects seriously.
That mixed outlook may describe many actual users. A writer can value faster research while worrying about reduced commissions. A programmer can use coding assistance while watching entry-level openings decline.
A manager can automate administrative tasks while questioning whether the organization will reinvest the savings in employees. An educator can use AI for preparation while worrying about student assessment.
These cases show why “concerned” should not be translated automatically into “anti-technology.” Concern can reflect informed use, direct exposure, or uncertainty about who controls the benefits.
The strongest conclusion remains narrow but important. Democratic concern has risen, conservative Republican concern has fallen, and job pessimism has spread across both parties.
Claims beyond that require additional evidence. The survey does not prove that Democrats will regulate AI, that Republicans will block every restriction, or that AI will eliminate a particular number of jobs.
Three Signals Will Show Whether the Reversal Lasts
The next test is whether concern becomes durable political behavior, concrete workplace policy, or another temporary shift in partisan mood.
The first signal is the next round of national polling. Researchers should track whether the Democratic-Republican gap persists after the midterms and whether general concern remains above 50%.
A sustained gap would suggest that AI has acquired a more stable partisan identity. A return to parity would indicate that the 2026 reversal reflected temporary political conditions.
The internal composition will matter as much as the headline number. Continued movement among liberal Democrats and conservative Republicans would strengthen the polarization thesis.
Broader movement among moderates would signal something larger. It would mean AI concern was spreading beyond the ideological groups currently driving the change.
The second signal is workplace evidence. Watch hiring plans, entry-level opportunities, job redesign, retraining commitments, and disclosures about automation.
Highly visible layoffs attributed to AI would reinforce AI job loss fears. Stable employment alongside measurable productivity gains would weaken the assumption that deployment necessarily reduces headcount.
Corporate language also deserves scrutiny. Companies often describe AI as assistance rather than replacement, but staffing decisions provide a clearer test than product marketing.
The important measure is not whether a tool can complete one task. It is whether employers change the number, type, and seniority of people they hire.
The third signal is federal and state policy. Concrete proposals will reveal whether either party can turn concern into an approach that survives internal disagreement.
For Democrats, the test involves connecting safety, labor protection, infrastructure, and accountability without collapsing them into one vague restriction. For Republicans, it involves reconciling rapid development with widespread expectations of job loss.
Data-center permitting offers an early test because it connects national AI ambitions with local costs. Workplace disclosure rules could become another, especially if automation affects hiring before unemployment data clearly change.
Model evaluations and incident reporting will shape the safety debate. Worker transition programs will shape the economic debate. These are related policy areas, but voters may judge them independently.
The Pew AI survey should therefore be read as an early warning, not an election forecast. It identifies a constituency becoming more worried and a second constituency that remains economically pessimistic.
AI companies cannot assume growing product use will produce political trust. Employers cannot assume efficiency claims will answer concerns about fewer jobs. Politicians cannot assume general anxiety supplies a ready-made policy.
For knowledge workers, the immediate task is to watch how employers use AI-derived productivity. Ask whether saved time expands better work, reduces hiring, or simply raises output expectations.
For business leaders, the question is whether adoption plans include transparent measures for job redesign and worker support. Silence leaves employees to assume the worst outcome.
For voters, the next stage requires moving beyond whether AI feels exciting or concerning. Which protections address documented harms, and which restrictions would merely signal action?
The partisan reversal will matter if it changes those decisions. Until then, it remains a clear measure of public anxiety and an unsettled warning about who Americans expect to benefit from AI.
