The Content Blind Spot
Most SaaS founders have the same content problem: they publish blog posts, check Google Analytics once a month, see flat traffic, and keep writing the same way. The gap isn't in writing ability — it's in data. Without continuous feedback between published content and SEO performance, every new post is a shot in the dark.
AIKit's Content Scoring Engine closes this loop. It reads your published posts, scores them across SEO dimensions, and suggests specific improvements — all without leaving the EmDash admin UI.
How the Scoring Engine Works
The Content Scoring Engine runs as an EmDash plugin inside a Cloudflare Workers sandbox. It uses a combination of LLM analysis and heuristic rules to evaluate each post across six dimensions:
1. Keyword Optimization (Score 0-100)
The engine extracts the primary keyword from your post title and first H2 heading, then checks:
- **Keyword density** — Is the target keyword used enough without keyword stuffing?
- **Semantic coverage** — Does the post cover related terms (LSI keywords) naturally?
- **Title tag inclusion** — Is the keyword in the title? (Critical for SEO)
2. Readability (Score 0-100)
Using the Flesch-Kincaid readability formula adapted for technical content:
```python
def readability_score(text):
words = len(text.split())
sentences = len([s for s in text.split('.') if s.strip()])
syllables = sum(count_syllables(w) for w in text.split() if w.isalpha())
if sentences == 0: return 50
score = 206.835 - 1.015 * (words / sentences) - 84.6 * (syllables / words)
return max(0, min(100, score + 20)) # Offset for technical content
```
Technical content naturally scores lower on readability (more jargon, longer sentences). The engine applies a +20 offset for technical blogs so you're not punished for using proper terminology.
3. Structure & Headings (Score 0-100)
The engine checks heading hierarchy:
- Does the post have at least one H2 heading? (Critical for scannability)
- Is the heading hierarchy logical? (No H3 before H2)
- Are headings descriptive? (Not just "Introduction" or "Overview")
- Is the post broken into digestible sections? (200-400 words per section)
4. Internal Linking (Score 0-100)
A post without internal links is an SEO island. The engine checks:
- Does the post link to other posts on the same site?
- Does it link to relevant category or tag pages?
- Are the anchor texts descriptive? (Not just "click here")
5. Freshness & Relevancy (Score 0-100)
Older posts lose SEO value. The engine tracks:
- Days since last update
- Whether the post references current data or outdated statistics
- Whether the URL structure is still valid
6. Mobile & Technical SEO (Score 0-100)
- Does the post use responsive images?
- Are code blocks properly formatted for mobile?
- Does the post load fast? (D1 queries in <5ms help here)
The Feedback Loop
Here's the flow that makes AIKit's approach different from a static SEO checker:
```
Publish Post → D1 Insert → Reader Views → Page View Tracking
↓
Content Scoring Engine ← Periodic Scan ← D1 Reads Data
↓
Score Report → LLM Suggests Edits → Auto-Update Draft Revision
↓
Editor Reviews → Approves Changes → Updated Post Goes Live
```
The key insight: the scoring engine runs on a schedule (daily), not on every page view. Each run scans the last 30 published posts and generates improvement suggestions. The editor reviews the suggestions and approves or rejects them. No spam, no unsolicited changes.
From 100 to 10,000: The Optimization Phase
At 100 published posts, we're entering the optimization phase. The first 100 posts built the foundation. The next 1000 will come from:
1. **Republishing old posts** — Adding new data, updating examples, refreshing SEO metadata
2. **Interlinking existing posts** — Creating topic clusters from the existing content graph
3. **Expanding thin content** — Posts under 800 words get expanded to 1,500+ words
4. **Consolidating overlapping posts** — Multiple posts on similar topics get merged
The Content Scoring Engine makes all of this data-driven. Instead of guessing which post to update next, the engine ranks all 100 posts by "optimization potential" — a weighted combination of current score and page view velocity.
Case Study: A Real Score Improvement
Here's an actual before-and-after from our dogfooding:
| Dimension | Before | After | Change |
|-----------|--------|-------|--------|
| Keyword Optimization | 62 | 88 | +26 |
| Readability | 71 | 83 | +12 |
| Structure | 55 | 92 | +37 |
| Internal Linking | 30 | 95 | +65 |
| Freshness | 45 | 90 | +45 |
| **Overall Score** | **52** | **90** | **+38** |
The fix was simple: add 3 internal links to related posts, break one long section into sub-sections with descriptive H3 headings, and update the title to include the primary keyword. The engine suggested all three changes automatically.
Conclusion
Building a content strategy without analytics is like sailing without instruments. AIKit's Content Scoring Engine gives every EmDash site a continuous, data-driven feedback loop between publishing and performance. At 100 posts, we have enough data to optimize. At 1,000 posts, the system will be recommending entire content clusters autonomously. The foundation is already in place — now it's about compound growth.