I Used AI Agents to Add 50+ Cross-Links to My Blog (And You Can Too)¶

I just had an AI agent read through 100+ of my blog posts and tell me exactly where to add internal links. In 30 minutes, it found connections I'd missed for years. Here's how I did it.

The Problem: Manual Cross-Linking Doesn't Scale¶

I have 100+ blog posts. Each one could potentially link to 5-10 others. That's 500-1000 potential cross-links.

Doing this manually means:

Reading every post (again)
Remembering what each one covers
Finding natural places to add links
Not making it feel forced

Or... I could have an AI agent do it in 30 minutes.

The Solution: AI Agent as Content Strategist¶

Here's my exact process:

Step 1: Generate AI Summaries¶

I have a Python script that runs GPT-4 on every blog post to create summaries. But the summaries aren't just for SEO – they're a map of my content.

Step 2: Feed Everything to an AI Agent¶

I gave Claude this prompt:

Here are summaries of all my blog posts.
Analyze them and tell me:
1. Which posts reference similar concepts
2. Where I mention a topic briefly that I cover in detail elsewhere
3. Natural places to add cross-links that would help readers

Step 3: Get Specific Recommendations¶

The AI agent returned specific, actionable suggestions:

"In consulting-start.md line 57, you mention pricing – link to consulting-pricing.md"
"Your RAG posts form a learning path – connect them sequentially"
"You reference hand injuries in 3 posts without linking to the full story"

Here's actual output from the AI agent:

File: consulting-start.md
Line ~89: "I learned this the hard way when I undercharged"
Suggestion: Link to money-negative-margin.md (discusses pricing mistakes)

File: rag-improving-rag.md
Line ~180: "I also wrote a 6 week email course on RAG"
Suggestion: Link to rag-course-breakdown.md (detailed course description)

File: ai-engineering-communication.md
Line ~110: "centered around RAG"
Suggestion: Link to rag-what-is-rag.md (explains RAG fundamentals)

50+ suggestions like this. All specific. All actionable.

The Implementation¶

The magic happens in generate_sitemap.py. Here's the core logic:

def should_regenerate_summary(file_path, file_hash, sitemap_data):
    """Check if we should regenerate the summary for a file"""
    relative_path = os.path.relpath(file_path, root_dir)

    # If file not in sitemap, generate summary
    if relative_path not in sitemap_data:
        return True

    # If hash changed, regenerate
    existing_hash = sitemap_data[relative_path].get('hash')
    if existing_hash != file_hash:
        return True

    # If no summary exists, generate
    if not sitemap_data[relative_path].get('summary'):
        return True

    return False

The beauty is in the two-step process:

Step 1: Summarize (Automated)

# This runs on every commit via pre-commit hook
for post in blog_posts:
    if content_changed(post):
        summary = gpt4_summarize(post)
        save_to_sitemap(summary)

Step 2: Find Links (AI Agent)

# This is where the magic happens
all_summaries = load_sitemap()
links = ai_agent_find_connections(all_summaries)
for link in links:
    print(f"{link.file}:{link.line} -> {link.target}")

The Results: 30 Minutes = 50+ Quality Cross-Links¶

In one session, the AI agent found:

15 links between consulting posts (creating a complete guide)
12 links in the RAG series (forming a learning path)
8 links from personal stories to relevant technical content
20+ links where I mentioned concepts briefly that I explain in detail elsewhere

Each suggestion included:

Exact file and line number
The text that should be linked
Which post to link to
Why the link adds value

Why AI Agents Are Perfect for This¶

AI agents excel at this because they can:

Hold 100+ posts in context simultaneously
Understand semantic relationships between different topics
Find non-obvious connections (like my hand injury story relating to why I started consulting)
Give specific line-by-line suggestions instead of vague advice
Work at superhuman speed (analyzing 100 posts in minutes)

The Implementation: Two Scripts, Endless Value¶

Script 1: generate_sitemap.py (runs automatically)

Generates AI summaries for changed blog posts
Maintains a cache to avoid regenerating everything
Runs on pre-commit hook
Costs: ~$0.01 per post

Script 2: AI Agent Analysis (run manually)

Feed all summaries to Claude/GPT-4
Ask for specific cross-linking opportunities
Get line-by-line suggestions
Implement the links

Total time: 30 minutes. Total cost: ~$2.

This Changes Everything¶

Think about what this means:

Every blogger can have perfect internal linking
Every content site can maximize reader engagement
Every knowledge base can be fully interconnected

And it takes 30 minutes.

This is what I mean when I talk about AI agents changing how we work. Not replacing us – augmenting us to do things we could never do manually.

Your Turn: Try This Now¶

Export your blog posts (or any content)
Feed them to Claude or GPT-4
Ask: "Find opportunities for cross-links between these posts"
Watch it find connections you missed

Or if you want to go full automation like me:

Set up automated summaries with GPT-4
Use those summaries for AI agent analysis
Never miss a cross-linking opportunity again

The future isn't AI writing our content. It's AI making our existing content work harder.

P.S. An AI agent helped me write this post by analyzing my commit history and identifying what made this approach unique. It suggested I focus on the agent aspect rather than the automation. Meta? Absolutely.