Prepared by Verasight

September Verasight National Survey

Survey results from a nationally representative sample of US adults aged 18+.

1,500 respondents
September 16–21, 2026
23 questions

Key Takeaways

AI adoption and pace
Americans want AI to slow down, and worry more about harm than about falling behind China
Data centers
Most oppose a data center nearby, but earmarked money flips them, most of all when it lowers their own bills
AI adoption and pace
Q1

AI Agent Familiarity

"AI agents" are artificial intelligence programs that can carry out tasks on their own — booking travel, making purchases, sending messages — without a person approving each step. Which best describes you?

(N = 1,500)
I have used an AI agent
29%
I know what they are, but have not used one
47%
I have heard the term, but am not sure what it means
13%
I have never heard of them
11%

Demographic Crosstabs

Response18-2930-4950-6465+
I have used an AI agent27%33%29%23%
I know what they are, but have not used one49%41%45%55%
I have heard the term, but am not sure what it means14%13%13%13%
I have never heard of them9%12%13%8%
Q2

Responsibility for an Agent Breach

Suppose a company uses an AI agent built by an outside technology company, and criminals exploit that agent to steal customers' personal information. How much responsibility, if any, does each of the following bear?

(N = 1,500)
A great deal
Some
Not much
None at all
Not sure
The company that was using the AI agent
A great deal
61%
Some
24%
Not much
7%
None at all
3%
Not sure
5%
The technology company that built the AI agent
A great deal
60%
Some
24%
Not much
7%
None at all
3%
Not sure
5%
The criminals who carried out the attack
A great deal
81%
Some
7%
Not much
5%
None at all
2%
Not sure
5%
Government regulators, for not setting clearer rules
A great deal
47%
Some
31%
Not much
12%
None at all
5%
Not sure
5%

Demographic Crosstabs

The company that was using the AI agent
Response18-2930-4950-6465+
A great deal54%53%66%72%
Some26%27%21%19%
Not much12%8%6%4%
None at all3%5%3%2%
Not sure6%7%3%4%
The technology company that built the AI agent
Response18-2930-4950-6465+
A great deal56%52%63%70%
Some27%26%22%21%
Not much7%10%8%5%
None at all3%4%4%2%
Not sure7%8%4%3%
The criminals who carried out the attack
Response18-2930-4950-6465+
A great deal73%76%84%92%
Some11%8%5%2%
Not much6%7%5%2%
None at all3%3%3%1%
Not sure6%7%4%3%
Government regulators, for not setting clearer rules
Response18-2930-4950-6465+
A great deal45%41%51%54%
Some32%30%30%31%
Not much13%14%11%8%
None at all5%7%5%4%
Not sure5%8%4%3%
Q3

Interest in Learning AI Tools

How interested are you, if at all, in learning more about how to use AI tools?

(N = 1,500)
21%
33%
23%
22%
Very interestedSomewhat interestedNot too interestedNot at all interested

Demographic Crosstabs

Response18-2930-4950-6465+
Very interested15%29%21%16%
Somewhat interested32%31%32%37%
Not too interested25%20%24%25%
Not at all interested28%20%22%21%
Q4

Ways People Would Learn AI Tools

Which of these would you be willing to do to learn more about using AI tools? Select all that apply.

(N = 1,500)
Watch a short video online, like on YouTube or TikTok
40%
Take a free online or in-person class
40%
Read articles or a book about it
33%
Learn from a friend or family member
32%
Ask an AI tool to teach you directly
31%
Take training offered by my employer
26%
None of these — I'm not interested
29%

Demographic Crosstabs

Response18-2930-4950-6465+
Watch a short video online, like on YouTube or TikTok40%44%40%36%
Take a free online or in-person class31%40%44%43%
Read articles or a book about it30%31%35%34%
Learn from a friend or family member29%29%35%35%
Ask an AI tool to teach you directly29%36%31%25%
Take training offered by my employer28%32%31%13%
None of these — I'm not interested32%28%28%31%

Percent selecting each option within the column; columns do not sum to 100.

Q5

Most Likely Way to Learn AI Tools

And which one are you most likely to actually do?

(N = 799; asked of respondents who selected two or more options in Q4)
Watch a short video online, like on YouTube or TikTok
30%
Take a free online or in-person class
20%
Ask an AI tool to teach you directly
17%
Learn from a friend or family member
11%
Take training offered by my employer
11%
Read articles or a book about it
10%

Demographic Crosstabs

Response18-2930-4950-6465+
Watch a short video online, like on YouTube or TikTok31%29%31%29%
Take a free online or in-person class13%24%22%20%
Ask an AI tool to teach you directly21%21%12%14%
Learn from a friend or family member12%5%12%18%
Take training offered by my employer10%13%13%7%
Read articles or a book about it13%7%11%12%
Q6

Awareness of AI Slowdown Calls

How much, if anything, have you heard or read about AI company executives recently saying the industry should slow down how fast it develops artificial intelligence?

(N = 1,500)
41%
37%
22%
A lotA littleNothing at all

Demographic Crosstabs

Response18-2930-4950-6465+
A lot39%35%42%49%
A little39%41%36%32%
Nothing at all22%24%22%19%
Q7

AI Harm vs. Falling Behind China

Which of these worries you more about artificial intelligence — AI could be developed too quickly and cause serious harm, or the United States could fall behind China in developing AI?

(N = 1,500)
AI could be developed too quickly and cause serious harm
45%
The United States could fall behind China in developing AI
11%
Both worry me equally
36%
Neither worries me
8%

Demographic Crosstabs

Response18-2930-4950-6465+
AI could be developed too quickly and cause serious harm58%46%41%37%
The United States could fall behind China in developing AI9%13%9%10%
Both worry me equally23%31%43%46%
Neither worries me10%10%7%6%
Q8

Who Should Slow AI, If Anyone

Some AI company executives have recently said the industry should slow down. Who should be responsible for slowing down the pace of AI development?

(N = 1,500)
55%
37%
8%
The companies themselvesThe governmentThe pace should not slow down

Demographic Crosstabs

Response18-2930-4950-6465+
The companies themselves52%53%59%56%
The government40%38%33%36%
The pace should not slow down8%9%8%7%
Q9

Preferred Pace of AI Development

Thinking about how fast artificial intelligence is being developed, which comes closest to your view?

(N = 1,500)
Development should slow down
63%
Development should continue at about the current pace
19%
Development should speed up
5%
Not sure
14%

Demographic Crosstabs

Response18-2930-4950-6465+
Development should slow down60%57%67%69%
Development should continue at about the current pace21%21%15%15%
Development should speed up7%4%5%2%
Not sure12%17%12%13%
Q10

Speed vs. Safety in AI Tools

Suppose two AI companies offer similar tools. One develops new features quickly. The other develops more slowly and spends more time on safety testing, so its tools are somewhat less advanced. Which would you choose?

(N = 1,500)
The company that develops quickly, with more advanced tools
12%
The company that develops slowly, with more safety testing
73%
Not sure
15%

Demographic Crosstabs

Response18-2930-4950-6465+
The company that develops quickly, with more advanced tools17%15%9%6%
The company that develops slowly, with more safety testing68%68%77%81%
Not sure15%17%14%13%
Q11

AI vs. Nuclear Weapons Risk

Thinking about the long term, which do you think poses a greater risk to humanity — AI or nuclear weapons?

(N = 1,500)
24%
23%
50%
4%
AINuclear weaponsBoth about equallyNeither poses a serious risk

Demographic Crosstabs

Response18-2930-4950-6465+
AI27%24%23%21%
Nuclear weapons25%23%19%24%
Both about equally44%48%54%54%
Neither poses a serious risk4%5%4%1%
The Meta settlement
Q12

Awareness of the Meta Settlement

How much, if anything, have you heard or read about the recent agreement between Meta and a group of state attorneys general over how Facebook and Instagram affect children?

(N = 1,500)
19%
55%
27%
A lotA littleNothing at all

Demographic Crosstabs

Response18-2930-4950-6465+
A lot20%20%17%17%
A little55%50%56%61%
Nothing at all24%30%28%22%
Q13

Expected Impact on Teen Safety

As part of the agreement, Meta must make several changes to Facebook and Instagram for users under 18, and pay up to $17 billion to the states over ten years. How much difference, if any, do you think this will make to the safety of teenagers using these apps?

(N = 1,500)
8%
31%
40%
15%
6%
A great dealSomeNot muchNone at allNot sure

Demographic Crosstabs

Response18-2930-4950-6465+
A great deal11%8%6%6%
Some39%29%25%34%
Not much32%39%44%43%
None at all14%16%18%11%
Not sure4%8%7%6%
Q14

$17B Penalty: Dollar Figure Only

Thinking about the $17 billion Meta agreed to pay over ten years, do you think that penalty…

(N = 762; asked of a random half of respondents; the other half saw Q15)
9%
29%
38%
23%
Went too farWas about rightDid not go far enoughNot sure

Demographic Crosstabs

Response18-2930-4950-6465+
Went too far6%8%10%12%
Was about right38%31%28%22%
Did not go far enough39%37%41%38%
Not sure18%24%21%28%
Q15

$17B Penalty: With Revenue Context

Thinking about the $17 billion Meta agreed to pay over ten years — roughly 1% of the company's annual revenue — do you think that penalty…

(N = 738; asked of the other random half of respondents; see Q14)
7%
21%
55%
16%
Went too farWas about rightDid not go far enoughNot sure

Demographic Crosstabs

Response30-4950-6465+
Went too far7%10%6%
Was about right21%17%17%
Did not go far enough57%51%60%
Not sure15%23%16%

Not shown: 18-29 (fewer than 100 respondents).

Q16

Favor/Oppose Settlement Remedies

The agreement requires Meta to make these changes for users under 18. Do you favor or oppose each one?

(N = 1,500)
Strongly favor
Somewhat favor
Somewhat oppose
Strongly oppose
Not sure
Limiting use to two hours a day by default
Strongly favor
42%
Somewhat favor
30%
Somewhat oppose
11%
Strongly oppose
7%
Not sure
10%
Blocking the apps between midnight and 6 a.m.
Strongly favor
52%
Somewhat favor
24%
Somewhat oppose
9%
Strongly oppose
6%
Not sure
9%
Removing "like" counts from posts by users under 18
Strongly favor
43%
Somewhat favor
27%
Somewhat oppose
10%
Strongly oppose
5%
Not sure
16%
Offering a version of the feed that is not personalized by an algorithm
Strongly favor
47%
Somewhat favor
27%
Somewhat oppose
7%
Strongly oppose
3%
Not sure
17%
Banning filters that change how a user's face looks
Strongly favor
48%
Somewhat favor
24%
Somewhat oppose
9%
Strongly oppose
6%
Not sure
13%

Demographic Crosstabs

Limiting use to two hours a day by default
Response18-2930-4950-6465+
Strongly favor33%44%41%47%
Somewhat favor33%26%31%32%
Somewhat oppose13%11%13%8%
Strongly oppose10%7%6%4%
Not sure11%11%9%8%
Blocking the apps between midnight and 6 a.m.
Response18-2930-4950-6465+
Strongly favor39%49%54%65%
Somewhat favor29%24%25%19%
Somewhat oppose14%10%9%4%
Strongly oppose7%7%3%4%
Not sure10%10%8%8%
Removing "like" counts from posts by users under 18
Response18-2930-4950-6465+
Strongly favor39%43%43%46%
Somewhat favor25%26%28%28%
Somewhat oppose13%10%10%6%
Strongly oppose7%6%4%3%
Not sure15%16%15%17%
Offering a version of the feed that is not personalized by an algorithm
Response18-2930-4950-6465+
Strongly favor42%48%49%47%
Somewhat favor30%25%27%26%
Somewhat oppose8%8%6%4%
Strongly oppose5%3%3%2%
Not sure15%16%15%21%
Banning filters that change how a user's face looks
Response18-2930-4950-6465+
Strongly favor40%45%53%54%
Somewhat favor26%26%21%22%
Somewhat oppose12%9%9%8%
Strongly oppose8%6%5%4%
Not sure15%13%11%12%
Q17

Extending Rules to Other Platforms

Do you think other social media companies — like TikTok, Snapchat, and YouTube — should have to make these same changes for their users under 18?

(N = 1,500)
Yes, they should be required to
67%
They should be encouraged to, but not required
21%
No, they should not have to
5%
Not sure
7%

Demographic Crosstabs

Response18-2930-4950-6465+
Yes, they should be required to48%65%74%78%
They should be encouraged to, but not required34%22%15%14%
No, they should not have to8%4%5%3%
Not sure10%9%6%4%
Data centers
Q18

Data Center Proximity

As far as you know, is there a large data center — a building full of computers that powers things like cloud storage and artificial intelligence — currently operating or planned within about 10 miles of your home?

(N = 1,500)
Yes, one is already operating
10%
Yes, one is planned or proposed
20%
No
40%
Not sure
30%

Demographic Crosstabs

Response18-2930-4950-6465+
Yes, one is already operating11%11%8%7%
Yes, one is planned or proposed24%20%18%18%
No40%37%37%46%
Not sure25%31%36%28%
Q19

Favor/Oppose a Nearby Data Center

Would you favor or oppose a large data center being built within 10 miles of your home?

(N = 1,500)
5%
17%
24%
53%
Strongly favorSomewhat favorSomewhat opposeStrongly oppose

Demographic Crosstabs

Response18-2930-4950-6465+
Strongly favor3%5%5%6%
Somewhat favor17%16%18%19%
Somewhat oppose20%24%26%26%
Strongly oppose59%55%51%48%
Q20

Data Center With a $100M Investment

Now suppose the company building that data center agreed to invest about $100 million in your community over ten years. To what extent would you favor or oppose the data center being built within 10 miles of your home if the money were used in each of these ways?

(N = 1,500)
Total favor
Strongly favorSomewhat favor
Somewhat opposeStrongly oppose
Total oppose
Paid to your local government to use however they wish26%
7
19
27
47
74%
$50,000 bonus for all public school teachers55%
22
33
19
27
45%
Paid to local police and fire departments53%
19
34
20
27
47%
Paid directly to residents, about $5,000 per household66%
36
30
13
22
34%
Used to lower residents' electricity bills70%
40
30
11
19
30%

Demographic Crosstabs

Paid to your local government to use however they wish
Response18-2930-4950-6465+
Strongly favor7%8%6%7%
Somewhat favor16%19%15%23%
Somewhat oppose30%24%26%29%
Strongly oppose46%50%53%40%
$50,000 bonus for all public school teachers
Response18-2930-4950-6465+
Strongly favor22%23%21%20%
Somewhat favor40%35%29%30%
Somewhat oppose16%17%19%22%
Strongly oppose22%25%30%29%
Paid to local police and fire departments
Response18-2930-4950-6465+
Strongly favor10%18%20%26%
Somewhat favor32%30%38%38%
Somewhat oppose27%21%17%16%
Strongly oppose31%30%25%20%
Paid directly to residents, about $5,000 per household
Response18-2930-4950-6465+
Strongly favor31%39%40%34%
Somewhat favor35%26%28%31%
Somewhat oppose13%12%11%14%
Strongly oppose21%22%21%22%
Used to lower residents' electricity bills
Response18-2930-4950-6465+
Strongly favor34%38%41%45%
Somewhat favor31%30%29%30%
Somewhat oppose14%11%12%9%
Strongly oppose20%21%18%16%
Q21

Willingness to Sell Property

Suppose you owned property — for example, a house/apartment, a building, or land — and a company that builds data centers offered you double what the property is worth. Would you sell?

(N = 1,500)
20%
35%
23%
22%
Definitely would sellProbably would sellProbably would not sellDefinitely would not sell

Demographic Crosstabs

Response18-2930-4950-6465+
Definitely would sell16%19%23%22%
Probably would sell33%35%34%36%
Probably would not sell25%22%21%24%
Definitely would not sell26%23%22%18%
Q22

Community Say on a Neighbor’s Sale

Now suppose someone near you received that offer and wanted to sell their property to a data center company. Which comes closest to your view?

(N = 1,500)
It is entirely their decision — neighbors should have no say
46%
The local community should be able to block the sale if most people object
54%

Demographic Crosstabs

Response18-2930-4950-6465+
It is entirely their decision — neighbors should have no say42%46%48%48%
The local community should be able to block the sale if most people object58%54%52%52%
The $5,000 midterm promise
Q23

Reported Effect of $5,000 Promise

President Trump has promised to send every adult American citizen $5,000 if Republicans keep control of both the House and the Senate in the midterm elections. Does this promise make you more or less likely to vote for a Republican in the midterm elections?

(N = 1,500)
15%
33%
52%
More likelyLess likelyNo change

Demographic Crosstabs

Response18-2930-4950-6465+
More likely19%19%13%10%
Less likely41%29%30%32%
No change39%52%57%58%

Survey Details

📋 1,500 US adults 📅 September 16–21, 2026

Verasight collected data for this project from September 16-September 21, 2026. The sample consists of 1,500 United States adults.

The sampling criteria for this survey were:

  1. US resident
  2. Adult (age 18+)

The selection criteria for the final sample were:

  1. Passed all data quality assurance checks, outlined below

The data are weighted to match the August 2026 Current Population Survey on age, race/ethnicity, sex, income, education, region, and metropolitan status, as well as to a running three-year average of partisanship distributions from the Pew Research Center NPORS benchmarking surveys and population benchmarks of 2024 vote. The margin of sampling error, which accounts for the design effect and is calculated using the classical random sampling formula, is +/- 2.6%.

All respondents were recruited via email from the Verasight Community, which is composed of individuals recruited via random address-based sampling, random person-to-person text messaging, and dynamic online targeting. All Verasight community members are verified via multi-step authentication, including providing an SMS response from a mobile phone registered with a major U.S. carrier (e.g., no VOIP or internet phones) as well as within-survey technology, including verifying the absence of non-human responses with technologies such as Google reCAPTCHA v3. Those who exhibit low-quality response behaviors over time, such as straight-lining or speeding, are also removed and prohibited from further participation in the community. Verasight Community members receive points for taking surveys that can be redeemed for Venmo or PayPal payments, gift cards, or charitable donations. Respondents are never routed from one survey to another and receive compensation for every invited survey, so there is never an incentive to respond strategically to survey qualification screener questions.

To further ensure data quality, the Verasight data team implements a number of post-data collection quality assurance procedures, including confirming that all responses correspond with U.S. IP addresses, confirming no duplicate respondents, verifying the absence of non-human responses, and removing any respondents who failed in-survey attention and/or straight-lining checks. The Verasight data team also reviewed open-ended items to ensure no responses contained nonsensical, inappropriate, or non-sequitur text. Respondents that completed the survey in less than 30% of the median completion time were removed.

Unmeasured error in this or any other survey may exist. Verasight is a member of the American Association for Public Opinion Research Transparency Initiative.

Reading the Crosstabs

Columns with fewer than 100 respondents are not shown. This report has 7 such columns; each affected table lists them in a “Not shown” note beneath it.

Crosstab figures are weighted column percentages. Respondents with no value on a banner variable are left out of that banner’s columns. Subgroup estimates carry wider margins of error than the full sample’s ±2.6 percentage points.

Percentages are rounded to whole numbers and may not sum to 100. Figures in the Key Takeaways and the Total favor/oppose columns of the matrix charts are computed from unrounded values, so a figure can differ from the sum of the displayed rows by one point.