Ask any audience anything.
userken builds synthetic audiences out of public data, then lets you put a question to them and read the answer in seconds, with the segment breakdown and the real quotes behind it. No panel to recruit, no survey data of your own needed to start.
Three steps, no fieldwork.
We build the audience from public data
Every public review we index gets embedded and clustered into personas with real weights: how many people actually sound like that. You pick an audience by category, by competitor, by app, or by filter. Nothing of yours is required to begin.
You ask, in the app or inside Claude
Single choice, multi select, Likert, ranking or open ended. Use the web app, or call the same tools over MCP so the question happens inside the Claude conversation you were already having.
Results in seconds, with the breakdown
A weighted distribution across the whole audience, the split by persona segment, the verbatims each segment gave, and the review ids that grounded the answer so you can go read the source.
Start with ours. Build your own.
Describe an audience in a sentence and userken turns it into filters, previews it, and builds personas for it. "Android users of financial news apps who complain about login" is a working audience in under a minute.
Real answers, not guesswork.
Concept testing
Put an early idea in front of a category audience before you spend a sprint on it.
Creative testing
Pre-read how creative work lands with the people who have to scroll past it.
Message testing
Find the sentence that does the work, and the one that quietly costs you users.
Pricing testing
Get a price range and the reasoning behind it before you touch the paywall.
A/B and A/B/n testing
Rank options against each other instead of scoring them one at a time.
Audience discovery
Find the segments that actually exist in a category, not the ones on a slide.
Competitive intelligence
Interview your competitor's users, because their reviews are public.
Focus groups
Moderate a conversation with five to eight personas and follow the thread.
The same tools, where you already work.
Every audience, survey and focus group is also an MCP tool. Connect userken to Claude with your API key and ask the audience mid conversation, or paste a focus group prompt and moderate the room yourself.
ask_audience(
question="Would you pay $4.99 a month to remove ads?",
options=["Yes", "No", "Not sure"],
audience_ref="cat:news",
samples_per_persona=5,
)
# weighted distribution, by_persona split,
# verbatims and grounding_review_ids
You can check our work.
We hold out recent reviews, ask the audience, and score the answer against what real people said. Two measures, published every run, next to the human ceiling and an ungrounded model.
Where it sits.
| userken | General LLM | Traditional research | |
|---|---|---|---|
| Time to an answer | Seconds to minutes | Seconds | Weeks |
| Grounded in real people | Public reviews, ids recorded | No | Yes, recruited |
| Segment breakdown | Weighted personas | One voice | Yes, if sampled |
| Accuracy method published | Yes, on a public page | No | Per study |
| Audience for a competitor | Yes, the data is public | Nothing specific | Hard and costly |
| Representative of a population | No, and we say so | No | Yes, when designed for it |
| Cost shape | Free tier, then credits | Your existing subscription | Per respondent |
Stop guessing. Start asking.
Free on the public audiences. Pro from $39 a month for your own.