The questions we get most often about how the focus group works, what you'll receive, and how we handle your data.
A persona is a chat-driven character with a goal, a backstory, and a set of preferences — a 28-year-old freelancer hunting for income, a busy product manager validating a launch, a skeptical retiree comparing options. We give each persona your URL and let them browse the way that character would: clicking, scrolling, getting lost, doubling back, leaving when something breaks their trust.
Personas are powered by frontier language models with structured prompts and a real headless browser. They don't roleplay survey answers — they actually click buttons and read what loads.
Most reports land in your inbox within 30 minutes. Larger jobs (the Crowd tier with 10 personas and a deep crawl) can take up to an hour. You'll get an email when it's ready, plus a status page you can refresh while it runs.
Direct quotes from each persona, per-persona ratings (ease of use, informativeness, ease of understanding, overall impression), an interaction log of every click and back-button, the questions visitors arrived with vs. the ones your site answered, friction points, broken trust signals, the elements nobody discovered, and behavioral conversion stats. The sample reports on the home page are the most honest answer to "what do I actually get?"
The default panels are tuned to your URL automatically — we infer audience from the site itself. For custom panels (specific industries, locales, professional segments, accessibility profiles), email [email protected] with what you need. We'll write back the same day.
Public-facing pages work out of the box. For auth-walled sites, staging URLs that need basic auth, or single-use share links, contact us — we have a custom-panel flow that supports them.
Today the report is delivered as a PDF only. Raw data export is on the roadmap; if you need it now, contact us and we'll arrange a one-off JSON dump.
Yes — run two tests with the same tier and compare the reports side-by-side. Native multi-variant testing in a single report is on the roadmap.
Real-user testing (UserTesting, Maze, recruited focus groups) gives you actual humans with actual context — the gold standard for accessibility nuance, brand resonance, and emotional reactions. It also costs $400–$40,000 and takes days to weeks.
AI personas are best for fast, cheap iteration: catching dead-end CTAs, broken value props, confusing copy, hidden navigation, and trust-breaking elements. Use us before, between, and around your real-user studies — not as a replacement for them when the stakes are high.
Subtle accessibility issues (motion sensitivity, screen-reader nuance), strong emotional reactions to brand or imagery, region-specific cultural cues, and anything requiring real-world context (like whether your pricing actually fits a household budget). We're transparent about this — the report is most useful when paired with occasional real-user studies for the high-stakes calls.
Full refund. If something breaks on our end — site couldn't be crawled, models timed out, PDF didn't render — we refund the charge automatically and tell you why. No back-and-forth.
Yes. If you're running tests on behalf of clients or buying 10+ tests per month, email [email protected]. We'll work out a rate that makes sense.
You're redirected to a status page that streams progress live. The pipeline kicks off within seconds of the Stripe webhook landing. When the report's ready, the status page shows a download link and we email a copy to the address you submitted.
Your URL is stored in our database alongside your report and visible only to your account (keyed to your email). We don't sell URLs, share them, or train models on them.
Reports are retained so you can re-download them later. To delete a report — or every record we have of you — email [email protected] from the address you submitted and we'll wipe it within one business day.
No. The personas use general-purpose foundation models that we don't fine-tune. Your site's content flows through the model as part of a single conversation; nothing about your URL or its content gets baked into a future model.