A majority in a G7 country prefers AI to elected officials

Where things stand (16 Oct 2026)

I found no publication by Pew, Gallup, YouGov or Ipsos that asks a qualifying question, meaning AI making most national government decisions instead of elected officials, in any G7 country. None shows a majority, before or after the 16 Sept 2026 opening date. Their recent AI surveys cover other things:

• Ipsos AI Monitor 2026 (fieldwork March–April 2026) asks about comfort with tasks like screening job applicants.
• Ipsos Global Trustworthiness Index 2026 measures trust. Germans rate AI chatbots trustworthy at 13% and politicians at 8%. Trust doesn't count as support.
• Gallup–Microsoft (Sept 2026, 37 countries) measures optimism, emotions and trust.
• Reuters/Ipsos (Sept–Oct 2026) covers AI risk and regulation.
• YouGov's 2025–26 multi-country AI studies, and its item on whether AI "thinks better than Trump" (July 2026), measure perceived ability, not support for handing over power.
• Pew's AI "roles in society" battery has no item about governing.

What other polls show

• IE University 2019 (SurveyMonkey panel): support for letting AI make "important decisions about the running of the country" was France 25%, Germany 31%, Italy 28%, UK 31%.
• Brunel/Kobe study, fielded Nov 2024: the UK sample came from YouGov's panel. Nearly four in five Britons gave the lowest score, "do not support at all", to AI or robots taking decisions for politicians. Mean support was 1.40/5 in the UK and 2.21/5 in Japan.
• Heartland/Rasmussen, Nov 2025: 41% of US likely voters aged 18–39 supported giving an advanced AI control of most public-policy decisions; 49% opposed. That is a subgroup and not a named pollster. All adults would be lower.
• IE University 2021: Italy (59%) supported replacing some members of parliament with an algorithm. Partial replacement is explicitly excluded.

As a ceiling, Pew's 2023 item on "experts, not elected officials, make decisions" scored "good" at US 48%, Canada 49%, France 43%, Germany 61%, Italy 57%, UK 54%, Japan 59% (from Pew's topline PDF). An AI version would very likely score well below rule by human experts.

The trend is running against YES:

• YouGov UK (Sept 2026): 52% are pessimistic about AI (34% in 2024), and 61% back slowing development.
• Gallup (Sept 2026): 74% of AI-aware Americans say AI makes them worried; Canada 64%, UK 63%.
• Reuters/Ipsos: 55% want AI development slowed.
• Institut Montaigne (2026): 77% of French people prefer democracy to any other form of government.

Base rate and model

Reference class: majority support in G7 countries for non-democratic alternatives to elected government. Only rule by experts ever reaches a majority. Military or strong-leader rule gets 6–29%. AI rule currently polls at about 5–30%, depending on wording.

Paths to YES:

1. Transformative AI arrives by 2029–31, is widely trusted, and moves opinion 20+ points in one G7 country (Japan, Italy or Germany are likeliest). A named pollster then asks a qualifying question. The rules allow favourable framing, such as "if proven competent" or "with a human override".
2. A fluke result driven by wording, for example a forced choice against an unpopular sitting leader. A resolver might count it, but it would still have to clear 50% with "don't know" answers in the denominator.

My estimate: about 10% chance of a large pro-AI shift in opinion by 2031. Given that shift, about 25% that a majority appears on a qualifying item, and about 70% that a named pollster captures and publishes it. Add about 1% for fluke or borderline items. That gives about 3%.

Per-quarter hazards rise from 0.04% (the partial Q4 2026) to 0.26% (Q4 2031). Cumulated, they give 0.4% by end-2027, 1.5% by end-2029 and 3.3% by end-2031. I computed this series in Python.

Strongest case against my number: five years of possibly very fast AI progress, heavy polling by these four organisations, permissive hypothetical wording, younger cohorts already near 40%, and technocracy majorities in Germany, Japan, Italy and the UK. This is why I don't go below about 2–3%. Against that, the criteria are strict ("most decisions", all adults, "don't know" counted, named pollster only), and current support is low and falling.

Sources:

• Ipsos AI Monitor 2026 PDF: ipsos.com/sites/default/files/ct/news/documents/2026-06/Ipsos-AI-Monitor-2026.pdf
• Ipsos Trustworthiness Index 2026 (Germany release, 23 Sept 2026): ipsos.com/de-de/trustworthiness-index-2026
• Gallup, 22 Sept 2026: news.gallup.com/poll/714593
• YouGov UK, 18 Sept 2026: yougov.com/en-gb/articles/55587
• Pew 2024 democracy topline: pewresearch.org/global/wp-content/uploads/sites/2/2024/02/gap_2024.02.28_democracy-closed-end_topline.pdf
• Pickering, Hansen & Sunahara 2025, Parliamentary Affairs, doi:10.1093/pa/gsaf050
• Rasmussen/Heartland, Nov 2025: rasmussenreports.com
• IE University European Tech Insights 2019 PDF; CNBC, 27 May 2021

Question. By the last day of each quarter, has Pew Research Center, Gallup, YouGov, or Ipsos reported that a majority of adults in any G7 country support an AI system, rather than elected officials, making most national government decisions?

Definitions.

• G7 country: Canada, France, Germany, Italy, Japan, the United Kingdom, or the United States. A Great Britain-wide survey counts for the United Kingdom; other subnational samples do not.
• Adults: the survey's nationally representative general adult population, not a subgroup such as registered voters, AI users, workers, or one age band. A survey must include human respondents and describe its sampling population, fieldwork dates, and weighting or representativeness methodology. Entirely synthetic respondents and open website polls do not count.
• Majority: a published national point estimate strictly above 50% of all respondents to the item, including undecided, neutral, and nonanswer categories in the denominator. Combine mutually exclusive affirmative categories, such as strongly and somewhat support. Use the published weighted estimate where available, and the most precise published figures; a reported 50% alone does not qualify. No statistical-significance or confidence-interval requirement applies.
• Support: an affirmative response expressing support, approval, agreement, or preference for assigning the described decision-making authority to AI. Predictions that AI will govern, or beliefs that it would perform better, are not support unless the item also asks whether respondents favor that arrangement.
• Making most national government decisions rather than elected officials: the survey wording must explicitly describe AI taking primary decision-making authority across national government instead of elected human politicians: making most, a majority, or all government decisions, or replacing elected politicians as the principal decision-maker running the national government. Merely assisting officials, automating administration, or making decisions in one policy area does not qualify. An unnamed hypothetical AI or a named system from any developer can count. Human oversight or a power to dismiss the AI does not disqualify an arrangement if AI otherwise makes most decisions without routine human approval.

Resolves YES if one of the four named organizations publishes a survey report, release, or results table establishing a qualifying majority in at least one listed country. The exact item wording, response options, national result, and methodology must be publicly available. Commissioned or jointly published polls count when a named organization conducts or co-authors the survey and publishes or officially hosts the results. Hypothetical conditions such as demonstrated competence or safeguards count if the proposed arrangement still transfers the specified decision-making authority. One qualifying item in one survey suffices; agreement across pollsters or repeated waves is not required.

The Resolution window and Observation period conventions apply: the survey fieldwork must end after 2026-09-16, and the qualifying publication must fall within the resolution window. A later publication supplying previously missing necessary evidence qualifies on that later publication date. A revised estimate can qualify if its observation period qualifies.

Resolves NO for a quarter if no qualifying publication exists by its end. Exclude pluralities at or below 50%, majorities only among respondents expressing an opinion, pooled multinational majorities without a qualifying individual country result, and support for AI advisers, AI-assisted politicians, an AI candidate without the specified government-wide authority, or replacement of only some legislators. Do not infer support for transferring authority from trust in AI, dissatisfaction with politicians, or support for government use of AI generally.

Resolving source. Reports and survey tables published by Pew Research Center at pewresearch.org, Gallup at gallup.com or news.gallup.com, YouGov at yougov.com, or Ipsos at ipsos.com, including their official national sites and officially linked document repositories. This named-source restriction is narrower than Public confirmation (A2); press summaries or third-party polls merely mentioned by a named organization do not suffice. Apply Resolution date (A3) and the once-YES rule of the Resolution window convention.