Your tests pass.
Users still fail.

Persona-driven testing validates user journeys, not features. AI vision analysis sees what automation misses.

npx personaspec init --typescript
$ npx personaspec run --persona trial-user ▸ Running: trial-user (Alex, PM evaluating product) screenshots/dashboard-first-look.png captured ⌛ AI analyzing screenshot... ✓ dashboard loaded in 1.8s ! confusion: "Multiple CTAs competing for attention" ✗ frustration: "Search icon not visible above fold" screenshots/feature-discovery.png captured ⌛ AI analyzing screenshot... ✓ found 'AI Analysis' button ○ note: "Button label could be clearer" ▸ Summary: 2 passed | 1 note | 1 confusion | 1 frustration

The testing blind spot

Login form submits successfully → Can a new user complete login before frustration?
Search returns results → Does a power user find results faster than competitors?
Filter applies correctly → Is the filter state obvious after page reload?
Data displays in table → Can a first-time visitor understand what they're seeing?

See it in action

Traditional test: checkout API returns 200. Pass. But what does the user see?

checkout-trust.spec.ts TypeScript 
/**
 * Persona: Jordan - First-time buyer
 * Concern: Is my payment safe? Did this work?
 */

test('post-checkout trust signals', async ({ page }) => {
  await completeCheckout(page, testCard);
  collector.screenshot(page, 'post-purchase');
});
AI Vision Analysis
! confusion: "No order confirmation visible. User cannot verify purchase succeeded."
○ note: "Receipt/order number not displayed. User has no proof of purchase."
✗ frustration: "Loading spinner still visible - did the payment go through?"
Screenshot: post-purchase.png

AI feedback types

Core categories for everything the vision model finds. Plus emotional journey tracking for human-realistic testing.

Type Meaning Example

Behavioral Realism: Track emotional journeys over time. A frustrated user who starts at 60% frustration and encounters three small issues may abandon, while a calm user would continue. Learn more →

Stop testing features. Test journeys.

Get started Read methodology Case study GitHub