See reports
Directional400 synthetic responses · United States · November 2026

Would you pay to pre-test marketing messages with a synthetic audience?

Before you spend on a real survey, is it worth running a synthetic audience first? We asked 400 synthetic respondents — grounded in real US Census microdata — whether they'd pay to pre-test marketing messages this way. The honest answer: the concept lands, but the price has to be right. A fast, directional read, not a real poll.

The question

“Would you pay to pre-test marketing messages with a synthetic audience before running a real survey?”

Maybe, depends on price54%
No, I prefer real surveys46%

Margin of error on the topline ±5 points (95% confidence, 400 responses).

Key findings
What we asked

We put the question "Would you pay to pre-test marketing messages with a synthetic audience before running a real survey?" to a synthetic audience of 400 people, grounded in US microdata (US Census ACS 1-year PUMS microdata).

The audience represents US adults.

What they said

The most common answer was "Maybe, depends on price" at 54% (±5 points).

Full distribution: Maybe, depends on price (54%) and No, I prefer real surveys (46%).

Differences between groups

By age group: "Maybe, depends on price" was strongest among 18-24-year-olds (65%) and weakest among 65+-year-olds (33%).

By education: "Maybe, depends on price" was strongest among the "Graduate degree" group (72%) and weakest among the "No high school diploma" group (31%).

By income: "Maybe, depends on price" was strongest among the "$50–75k" group (70%) and weakest among the "<$25k" group (44%).

What it means

The price objection is the real barrier — not the concept. 54% are open to it, 0% said 'yes, definitely' without a caveat. That means a free first run or a very low entry price removes the main friction.

Younger audiences (18–44) are most receptive. If you're a marketer who wants to kill bad ideas cheaply before spending on real fielding, synthetic audience pre-testing is built for exactly that.

How honest this is

This is a synthetic study, not a real survey. Answers come from an AI audience grounded in US real microdata (age, sex, education and income jointly) (US Census ACS 1-year PUMS microdata). The distribution has genuine spread (2 options with meaningful support), but still treat it as a quick directional read, not a measured fact.

For high-stakes calls, use this as a first direction and validate with real people before you commit.

How “Maybe, depends on price” varied between groups
By age group
65+33%
35-4465%
25-3465%
55-6449%
45-5460%
18-2465%
By sex
Men56%
Women52%
By education
High school50%
Bachelor degree61%
Some college52%
No high school diploma31%
Graduate degree72%
Associate degree61%
By income
<$25k44%
$25–50k53%
$50–75k70%
$75–100k62%
$100–150k67%
$150k+60%
How honest this is — directional

This is a synthetic study, not a real survey. Answers come from an AI audience grounded in US real microdata (age, sex, education and income jointly) (US Census ACS 1-year PUMS microdata). The distribution has genuine spread (2 options with meaningful support), but still treat it as a quick directional read, not a measured fact.

Frequently asked questions

Is this a real survey of marketers?

No. It's a synthetic study — answers come from an AI audience grounded in US Census PUMS microdata (age, sex, education and income), not real respondents. Treat it as a quick directional read.

Why did nobody say 'yes, definitely'?

Because the right answer for most people is 'it depends on price and use case' — which is exactly the right positioning for a pre-testing tool. The concept is understood; the barrier is economic.

How do I try it myself?

The tool is free to try with no signup required. Ask any question, define your audience, and get a weighted response distribution in about 30 seconds.

Synthetic Audience Simulator — AI Survey on Real Census Data

Ask any question to a representative synthetic audience and get a simulated response distribution — grounded in real census microdata.

Method: 400 synthetic respondents, gpt-4o-mini, grounded in US microdata (US Census ACS 1-year PUMS microdata). Published November 2026. Generated with Chorus.