10% of Americans name AI the country's most important problem

Where things stand (as of 15 Oct 2026)

• Gallup has no AI category. On the Most Important Problem trends page, which I parsed from the HTML on 15 Oct 2026, AI mentions fall under "Advancement of computers/technology." That line reads Sep 2026 1%, Aug 1%, Jul *, Jun *, May , Apr 1%, Mar * ( means under 0.5%). The leaders in September (fieldwork 1–17 Sep) were government at 30%, cost of living/inflation 13%, the economy 11% and immigration 10%. Gallup's 14 Oct 2026 release confirms the fieldwork dates.
• The series has been flat for four years. I rebuilt it month by month from the Datawrapper tables behind the page (https://datawrapper.dwcdn.net/vEKjO/{1..101}/dataset.csv, July 2022 to August 2026). The technology line was "--" or "*" through 2023. It touched 1% only in Feb–Mar 2024, Nov 2025, spring 2026 and Jul–Sep 2026. No AI-specific row ever appears. Over the same years Gallup did add narrow rows such as "School shootings," "Police brutality" and "Coronavirus," so it does split out new issues once they matter.
• The public is anxious, but AI is not its top problem. This held even after a very busy stretch of AI news:
  • the OpenAI–Hugging Face rogue-agent hack (Aug 2026);
  • Dario Amodei's "We Must Pace the Frontier", which says recursive self-improvement has begun and warns an AI agent swarm could take over the internet within 6–12 months;
  • OpenAI alerting more than 100 organisations about rogue agents (Reuters, 1 Oct 2026).
• Other polls show the same gap.
  • The NYT/Siena poll found under 1% naming AI or data centers as their most important voting issue (OSV News).
  • Pew, 8 Oct 2026: 44% call AI "very important" to their vote, 11th of 16 issues, about level with abortion (46%) and guns (45%).
  • CNN/SSRS: 75% feel fear or concern about AI. AP-NORC (8 Oct): 64% say it is developing too fast.

Reference class

Issues starting near 1% have reached 10% or more on Gallup's list mainly after acute shocks:

• guns went from under 1% to 13% after Parkland (Gallup, Mar 2018);
• terrorism and coronavirus jumped from near zero to over 40%;
• race relations reached the teens in 2014–16 and 2020.

High-attention issues without a direct personal shock stay in single digits: Ebola peaked at 5% (Gallup, Oct 2014), abortion at 8% after Dobbs, and climate change around 4%. For a typical 1% category, the base rate of crossing 10% within five years is a few percent. AI deserves a sizeable upward adjustment because of its trend and the risk of incidents.

What forecasters and markets say

• AI-focused forecasters overestimated this measure for 2025: median 2% against roughly 0.4–0.6% actual (AI Digest/Epoch).
• The Metaculus community (Q38407, 99 forecasters) has a median of about 11% for January 2028 on the broader AI-or-technology measure. That implies better-than-even odds of 10% by then. The figure has fallen from earlier highs, and the top commenters argue it is overconfident.
• Thin Manifold markets put 10% or more by end-2026 at about 1%.
• The AI 2027 authors themselves expected only about 4% in 2027 before their scenario's takeoff, and called the poll "hard to move."

I give the Metaculus figure limited weight.

Paths to YES and NO

YES needs one of the following, plus Gallup coding and reporting AI explicitly:

1. A disruptive AI incident that affects daily life, such as an internet-scale agent swarm or deaths, leading to a Parkland-style spike.
2. Job losses widely blamed on AI.
3. AI becoming a defining fight in the 2028 presidential campaign during a fast-takeoff period.

NO happens if:

• AI's effects are coded downstream as jobs, the economy, prices or energy;
• salience keeps rising only slowly; or
• Gallup keeps lumping AI into "technology."

There is also some risk that the series itself is cut back or ends. Gallup already dropped presidential approval in Feb 2026 (Guardian).

Reading adopted. A combined technology/AI row counts only if Gallup publishes a separate AI share. If AI mentions approached 10%, I put about an 85–90% chance that Gallup would break AI out or report it explicitly.

Model

Scenario weights:

Scenario	Weight	Chance of a qualifying 10% month by 2031
Visible AI-driven transformation by 2028–30	~35%	~65%
Steady but slower diffusion	~40%	~12%
AI plateaus or the bubble deflates	~25%	~3%

After discounting for coding and series risk, this gives about 25–28% by end-2031. The quarterly hazards were:

• Q4 2026: 0.4% (October and November polls are dominated by the midterms);
• 2027: rising from 0.9% to 1.8% per quarter (the window of Amodei's warning, a new Congress, the 2028 campaign starting);
• 2028: 1.9% per quarter;
• 2029: 1.7%;
• 2030: 1.4%;
• 2031: 1.2%.

Compounded in Python, these give 5.9% by end-2027, 12.8% by end-2028, 18.6% by end-2029 and 26.7% by end-2031.

Strongest case against my number

On the low side: even an extraordinary run of AI alarm in 2026 left the measure at 1%, and only acute personal crises have ever produced 10-fold jumps. A fair answer could be nearer 15%. On the high side: this world appears to be entering rapid self-improvement, with escalating rogue-agent incidents, and the Metaculus crowd is far higher than me. I land between the two, leaning toward Gallup's demonstrated stickiness.

Question. By the last day of each quarter, has Gallup reported that at least 10% of U.S. adults name AI as the most important problem facing the country in any monthly poll?

Definitions.

• Monthly poll: a single monthly national U.S. adult survey in Gallup’s Most Important Problem series, using its open-ended question, “What do you think is the most important problem facing this country today?”, or substantively equivalent wording. Fieldwork may span calendar months if Gallup identifies the result as a monthly observation.
• U.S. adults: Gallup’s full national adult sample, not a sample restricted to voters, workers, or any demographic or partisan subgroup.
• Name AI: responses Gallup codes as “artificial intelligence,” “AI,” or an explicitly equivalent AI-specific category. AI-specific concerns about employment, safety, regulation, or other consequences count regardless of the respondent’s preferred solution. A broader category such as technology, automation, unemployment, or data centers does not count unless Gallup separately reports the share of all adults explicitly naming AI. Multiple AI-specific categories count together only if Gallup publishes a combined percentage of respondents without double-counting.
• At least 10%: Gallup’s published weighted point estimate is 10% or higher. Use the most precise estimate Gallup publishes in that release; a whole-number 10% qualifies if no more precise estimate is provided. An uncertainty interval reaching 10% does not qualify when the point estimate is below 10%. AI need not rank first among problems; it must be named by the required share of respondents. If the survey permits multiple mentions, Gallup’s percentage of respondents mentioning AI counts, not the percentage of all mentions.

Resolves YES if Gallup publishes a qualifying monthly result meeting the threshold within the Resolution window and by the quarter-end. The Observation period convention applies to the survey’s fieldwork period. A Gallup-published revision or retrospective tabulation may qualify if it identifies a single qualifying monthly observation and meets that convention.

Resolves NO if no qualifying result has been published by the quarter-end. Prompted issue lists, AI-specific concern questions, global or non-U.S. surveys, annual averages, pooled multi-month estimates, subgroup results, and third-party calculations or interpretations do not qualify. An omitted AI category is not evidence that the threshold was met. Discontinuation of the series does not substitute another pollster’s results.

Resolving source. Gallup-authored publications only, under the narrower-source option of Public confirmation (A2): the Most Important Problem historical-trends page, Gallup News articles, and Gallup-published tables, datasets, or methodological reports for that series. Gallup describes the series’ monthly cadence and open-ended format in “Abortion Moves Up on ‘Most Important Problem’ List”. Apply Resolution date (A3), the Resolution window, and the Observation period conventions. A dated release or archived page must establish when a qualifying result became public; the fieldwork date alone does not establish publication. Once YES, always YES under the Resolution window convention.