· 8 min read
Introducing Intent-Aware Test Generation & Coverage Recommendations
AegisRunner Team·April 3, 2026
Testing a shopping site and a blog are fundamentally different. We built intent-aware test generation that understands what your website IS and generates tests based on what real users would actually do.
The Problem
Our AI test generator was element-driven — it found buttons and links, then generated generic tests. Technically correct but missing the point.
✗ Before
Click button → assert visible
Fill form → assert exists
Click link → assert heading
Fill form → assert exists
Click link → assert heading
✓ After
Add to cart → verify badge count
Change quantity → verify subtotal
Submit comment → verify it appears
Change quantity → verify subtotal
Submit comment → verify it appears
The 4-Pass Pipeline
🔍
Pass 0
Classify
Detect site type
📝
Pass 1
Plan
Intent scenarios
🤖
Pass 2
AI Write
Playwright steps
✅
Pass 3
Validate
Fix selectors
Site Classification
After every crawl, we automatically detect your website type using URL patterns, button text, form types, and API endpoints.
🛒
E-commerce
Cart, checkout, products
📝
Blog / Content
Articles, comments, posts
💻
SaaS / Dashboard
CRUD, settings, users
👥
Social
Feed, likes, follows
✈️
Booking / Travel
Dates, reservations
🚀
Marketing
CTAs, pricing, demos
📌 Where to find it: Project → Settings → AI Configuration
Coverage Recommendations
After each crawl, we compare what was found vs. what a typical site of your type should have.
77%
Coverage Score
E-commerce • 29 pages • 1382 states
✓ Covered
Product pages
Add to Cart
Cart
Checkout
Search
Categories
Login
⚠ Missing
Reviews (low)
Wishlist (low)
📌 Where to find it: Project → Recommendations tab • Crawl → Recommendations tab
Other Improvements
🎯
Stable Selectors
data-testid and aria-label over brittle nth-of-type
🔄
Dynamic Auto-Fix
Cart badges use partial-match selectors automatically
🔓
Robots.txt Disabled
Testing tools crawl freely by default
🗃
Deduplication
Same page with different query params = one test suite
How to Use It
1
Crawl your site — classification happens automatically
2
Check detected type — Project → Settings → AI Configuration
3
View recommendations — Project → Recommendations tab
4
Run tests — they now match real user behavior
aitestingreleaseintent-detectioncoverage