Synthetic Audiences is best at narrowing a long list down to the few options worth testing with real people. Each example ends by asking what to confirm with a human survey. Keep that line in when you edit them.
Before you start
1
Open the PickFu Agent or your MCP client
Synthetic Audiences is in beta and open to all PickFu users. Run it in the PickFu Agent or through the PickFu MCP.
2
Copy a prompt
Find the example closest to what you’re testing and use the copy button.
3
Paste and edit
Replace each bracketed placeholder, such as
[your product] or [Title A], with your own details. Add or remove option lines to match how many options you have. For image examples, attach images in the PickFu Agent. In other MCP clients, follow the PickFu-hosted upload link your assistant gives you. Before uploading, verify that the URL’s hostname is app.pickfu.com. Keep the audience description and the closing line.E-commerce
Narrow down your main images
Got more images than you can afford to test? Find the few worth putting in front of real shoppers.Compare product titles
Find out which listing title earns the click. This is the kind of text comparison Synthetic Audiences handles best.Test benefit claims
Find out which promise actually lands, and why. Use it for bullet points, image callouts and ad headlines.Choose which prices to test
See how people react to different price points and offers before you test them for real.Check a new product idea
Get open-ended first reactions to an idea, and what would stop someone buying it.Publishing
Narrow down book covers
Find the cover concepts worth testing with real readers.Shortlist book titles
Too many title ideas to pay for? Cut the list to your finalists.Mobile games
Narrow down character designs
Find the characters players most want to play as.Shortlist game names
Cut a long list of name ideas down to the few worth testing with real players.What you get back
Results come back in the chat in a few minutes. For comparisons, you get the themes that drove the decision across your options, then a stronger half and a weaker half. You won’t see a vote split: synthetic respondents agree with each other more than real people do, so a precise-looking percentage would overstate how confident the result is. Use the direction, not the gap. For open-ended questions, you get the main themes with quotes from individual synthetic respondents. You can ask to see every answer.What to change, and what to keep
Change the audience description
Change the audience description
Describe who you want in plain language. PickFu picks the best-fitting synthetic audience, runs the test, and tells you which one it used. To try a different audience, ask for a new test with it. If none fits, it explains why the test can’t run. All audiences are U.S. only.
Change the options
Change the options
Any of these examples works for other things you need to choose between: logos, brand names, packaging, taglines or ad copy. Two options gives a head-to-head, and three or more are split into a stronger and a weaker half. Up to 10 works for screening a long list.
Text works best, images work too
Text works best, images work too
Synthetic Audiences is most reliable on text: names, titles, claims and price framing. It is less reliable on images because visual nuance is lost, so use image results as a hint and confirm aesthetic choices with a human survey.
Keep the sample size sensible
Keep the sample size sensible
30 to 50 is the recommendation and 75 is the cap. Past 50, synthetic respondents converge and extra responses add very little.
Keep the closing line
Keep the closing line
Asking which options to confirm with a human survey is the point. Synthetic decides what to test; real people decide what to do.
