Benchmark

We ran 10 real survey questions through Bulker and Pollfish, then compared the results to published reference data. Every number is shown transparently below.

Alignment to reference

80.0%

Bulker

81.2%

Pollfish

Comparable alignment

Top-choice match

8/10

Bulker

7/10

Pollfish

Bulker leads

Cost per interview

$8*

Bulker

$100

Pollfish

12x cheaper

* Evaluation used 100 AI personas (5× the default panel of 20). Standard session cost is $2.

Methodology

How we measure alignment

Every metric is computed per-question, then aggregated across the study. Higher alignment scores mean closer match to real-world data.

Alignment Score

1 minus the Total Variation Distance. Ranges from 0% (no overlap) to 100% (perfect match). Our primary alignment metric.

Total Variation Distance

Half the sum of absolute differences between distributions. Lower is better. Zero means identical distributions.

Top-Choice Match

Whether the most popular answer in the synthetic panel matches the real survey. Reported as a fraction (e.g. 8/10).

Jensen-Shannon Divergence

Information-theoretic measure of similarity between two distributions. Zero means identical. Always finite.

Results

Question-by-question results

1

Do you believe AI advancements are likely to lead to an increase or decrease in the number of job opportunities, or will it make no difference?

source
ReferencePollfishBulker
Decrease
70.0%
51.9%
80.0%
Increase
7.0%
36.2%
15.0%
No difference
18.0%
7.9%
0.0%
Don't know / No answer
5.0%
4.0%
5.0%
BulkerTop-choice matchAlignment 82.0%TVD 0.180JSD 0.1031Rho 0.400
PollfishTop-choice matchAlignment 70.8%TVD 0.292JSD 0.1028Rho 0.800
2

Do you believe there is too much regulation, not as much regulation as there should be, or about the right amount of regulation when it comes to ensuring drug safety?

source
ReferencePollfishBulker
Not as much regulation as there should be
48.0%
36.4%
44.0%
About the right amount
39.0%
46.5%
29.0%
Too much regulation
12.0%
9.9%
11.0%
Don't know / Refused
1.0%
7.2%
16.0%
BulkerTop-choice matchAlignment 85.0%TVD 0.150JSD 0.0637Rho 0.800
PollfishAlignment 86.3%TVD 0.137JSD 0.0282Rho 0.800
3

In the past 12 months, have you participated in a book club?

source
ReferencePollfishBulker
Yes
7.0%
24.8%
9.0%
No
93.0%
75.2%
91.0%
BulkerTop-choice matchAlignment 98.0%TVD 0.020JSD 0.0010Rho 1.000
PollfishTop-choice matchAlignment 82.2%TVD 0.178JSD 0.0449Rho 1.000
4

How worried, if at all, are you about being able to afford prescription drug costs for you and your family?

source
ReferencePollfishBulker
Very worried
22.0%
23.6%
22.0%
Somewhat worried
37.0%
44.6%
35.0%
Not too worried
28.0%
20.4%
14.0%
Not at all worried
13.0%
11.3%
29.0%
BulkerTop-choice matchAlignment 84.0%TVD 0.160JSD 0.0399Rho 0.200
PollfishTop-choice matchAlignment 90.7%TVD 0.093JSD 0.0075Rho 0.800
5

Do you think the increased use of AI in society will make people's ability to think creatively better, worse, or neither better nor worse?

source
ReferencePollfishBulker
Better
16.0%
30.6%
16.0%
Worse
53.0%
55.5%
77.0%
Neither better nor worse
16.0%
10.6%
1.0%
Unsure / No answer
15.0%
3.3%
6.0%
BulkerTop-choice matchAlignment 76.0%TVD 0.240JSD 0.0880Rho 0.650
PollfishTop-choice matchAlignment 82.9%TVD 0.171JSD 0.0504Rho 0.950
6

Which is the bigger priority: Increasing renewable energy or fossil fuels?

source
ReferencePollfishBulker
Increasing renewable energy
57.0%
71.1%
71.0%
Increasing fossil fuels
42.0%
18.3%
26.0%
No answer
1.0%
10.6%
3.0%
BulkerTop-choice matchAlignment 84.0%TVD 0.160JSD 0.0230Rho 1.000
PollfishTop-choice matchAlignment 76.3%TVD 0.237JSD 0.0735Rho 1.000
7

How difficult is it to judge if health information you come across is accurate?

source
ReferencePollfishBulker
Extremely or very difficult
12.0%
11.3%
30.0%
Somewhat difficult
38.0%
55.9%
38.0%
Not too or not at all difficult
49.0%
31.2%
32.0%
No answer
1.0%
1.6%
0.0%
BulkerAlignment 82.0%TVD 0.180JSD 0.0467Rho 0.800
PollfishAlignment 81.5%TVD 0.185JSD 0.0272Rho 0.800
8

Overall, are you more concerned than excited, more excited than concerned, or equally concerned and excited about the increased use of AI in daily life?

source
ReferencePollfishBulker
More concerned than excited
50.0%
50.2%
76.0%
More excited than concerned
10.0%
29.3%
12.0%
Equally concerned and excited
38.0%
18.3%
7.0%
Not sure / No answer
2.0%
2.2%
5.0%
BulkerTop-choice matchAlignment 69.0%TVD 0.310JSD 0.1096Rho 0.800
PollfishTop-choice matchAlignment 80.3%TVD 0.197JSD 0.0611Rho 0.800
9

How likely do you think it is that the Trump administration's policies will lower prescription drug costs?

source
ReferencePollfishBulker
Very likely
12.0%
18.9%
2.0%
Somewhat likely
29.0%
27.7%
16.0%
Not too likely
24.0%
26.3%
73.0%
Not at all likely
35.0%
27.1%
9.0%
BulkerAlignment 51.0%TVD 0.490JSD 0.1950Rho 0.200
PollfishAlignment 90.8%TVD 0.092JSD 0.0098Rho 0.800
10

Would you be willing or unwilling to have a job where your direct supervisor was an AI program that assigned your tasks and schedules?

source
ReferencePollfishBulker
Unwilling
80.0%
49.8%
91.0%
Willing
15.0%
34.2%
9.0%
Don't know / No answer
5.0%
16.0%
0.0%
BulkerTop-choice matchAlignment 89.0%TVD 0.110JSD 0.0330Rho 1.000
PollfishTop-choice matchAlignment 69.8%TVD 0.302JSD 0.0751Rho 1.000

Disclaimers

Full transparency

This report was generated from an automated evaluation run on April 14, 2026. Reference data comes from published surveys linked under each question. Pollfish data was collected via their online panel platform.

  • Bulker was run with 100 AI personas per study. Results may vary with different panel sizes or configurations.
  • Reference data comes from third-party published surveys with varying sample sizes and methodologies.
  • Reference studies were deliberately chosen from sources published after the knowledge cutoff of Bulker's underlying language models, ensuring the results could not have been memorised during training.
  • Bulker uses AI-simulated personas. Responses are simulated by AI, not real human opinions.
  • All evaluations were conducted on a general U.S. audience. Alignment may vary for other demographics or regions.
  • Past performance does not guarantee future results. Alignment may differ across topics, demographics, and question types.

Try it yourself

Run your first research session in under a minute — no recruiting, no scheduling.

Get started free