Imported from chinkan/aeo-score-chrome-ext (
AGENTS.md). Install upstream withnpx skills add chinkan/aeo-score-chrome-ext. Copyright stays with the author.
PROJECT KNOWLEDGE BASE — AEO Score Calculator
Generated: 2026-04-01 Project: Chrome MV3 Extension — Answer Engine Optimization scoring Status: Planning phase (no source code scaffolded yet)
OVERVIEW
Chrome extension that calculates AEO scores for webpages. Uses local LLM (@xenova/transformers) for tone/uniqueness analysis, heuristic scoring for EEAT/relevance/structure/freshness.
STRUCTURE
├── AGENTS.md # This file
├── package.json # [PLANNED]
├── vite.config.ts # [PLANNED]
├── manifest.json # [PLANNED] Chrome MV3 manifest
├── src/
│ ├── background/ # Service worker + message routing (DOM-free)
│ ├── content/ # DOM extraction + calculateAEO
│ ├── offscreen/ # Transformers.js execution
│ ├── popup/ # React UI (ScoreRing, Breakdown, Suggestions)
│ ├── lib/
│ │ ├── aeo-calculator.ts # 19 scoring functions
│ │ ├── aeo-scoring.ts
│ │ ├── llm/
│ │ └── messaging/
│ └── components/ui/ # shadcn components
├── tests/ # Vitest tests
└── dist/ # Built extension
WHERE TO LOOK
| Task | Location | Notes |
|---|---|---|
| Service worker | src/background/ |
DOM-free, message routing only |
| DOM analysis | src/content/ |
Main content extraction, 19 scoring funcs |
| LLM inference | src/offscreen/ |
Transformers.js (multilingual BERT sentiment + MiniLM) |
| Popup UI | src/popup/ |
React 19, ~380×520px |
| Scoring logic | src/lib/aeo-calculator.ts |
EEAT, Relevance, Structure, Freshness |
| Types/interfaces | src/lib/ |
AEOResult, Suggestion, component scores |
CONVENTIONS
- TypeScript strict mode
- Tailwind CSS v4 + shadcn/ui
- Vite + CRXJS / vite-plugin-web-extension
- Typed messaging between all components
ANTI-PATTERNS
- Background service worker must be DOM-free — no
document,window, browser globals - LLM code must run in offscreen document or via dynamic import
- Content script: limit DOM scan to
main,article,[role="main"] - Never suppress
document is not definederrors — fix root cause - No
as anyor@ts-ignore
SCORING FORMULA
final_score = (EEAT × 0.40) + (Relevance × 0.30) + (Structure × 0.20) + (Freshness × 0.10) + LLM bonus
19 CORE FUNCTIONS
| # | Function | Purpose |
|---|---|---|
| 1 | fetch_html() |
Get page HTML |
| 2 | extract_main_text() |
Clean main content (strip nav/footer/ads) |
| 3 | extract_schema_markup() |
JSON-LD + Microdata (FAQPage, HowTo, Article) |
| 4 | extract_meta_tags() |
Title, description, author, dates |
| 5 | get_last_updated() |
Date object from meta |
| 6 | has_faq_section() |
Detect FAQ section/schema |
| 7 | has_howto_schema() |
Detect HowTo schema |
| 8 | calculate_EEAT() |
0-1, weight 40% — author, citations, trust |
| 9 | calculate_Relevance() |
0-1, weight 30% — content vs userIntent |
| 10 | calculate_Structure() |
0-1, weight 20% — headings, lists, tables |
| 11 | calculate_Freshness() |
0-1, weight 10% — age-based decay |
| 12 | calculate_Intent_Match() |
0-1 — content matches intent |
| 13 | calculate_Tone() |
0-1 — multilingual 5-star sentiment mapped to tone |
| 14 | detect_snippet_ready() |
0-1 — snippet readiness |
| 15 | calculate_Uniqueness() |
0-1 — LLM embedding similarity |
| 16 | generate_suggestions() |
Actionable improvements |
| 17 | clamp() |
Number clamping utility |
| 18 | average() |
Array average utility |
| 19 | finalizeAEO() |
Combine heuristic + LLM scores → AEOResult |
AEOResult STRUCTURE
score: number(0-100)components: { eeat, relevance, structure, freshness, tone, uniqueness, intent_match }suggestions: Suggestion[]meta: { lastUpdated, hasFaq, hasHowto, ... }
COMMANDS
bun run dev # Development
bun run build # Production build (always verify after changes)
bun run test # Vitest testsNOTES
- Popup size: ~380 × 520px
- Freshness decay: <30d=1.0, <90d=0.7, <180d=0.4, older=0.1
- All 7 phases marked "Done" in plan but no source code exists yet — project needs scaffolding