I've been meaning to audit my website's SEO for a while now. You know how it goes—you build something, you deploy it, and then you move on to the next thing. But eventually, you realize that maybe, just maybe, you should check if search engines can actually find and understand your content properly.
So this morning, I fired up Screaming Frog SEO Spider and let it crawl my entire site. If you haven't used it before, Screaming Frog is this fantastic tool that basically acts like a search engine crawler. It goes through your site, checks for broken links, missing meta tags, duplicate content, all that good stuff. And it generates these detailed CSV reports that tell you exactly what's wrong.
The Results Were... Not Great
Let me just say, the results were eye-opening. The tool found:
- 40 internal 4xx errors (broken links and missing assets)
- 7 pages missing H1 tags (not great for SEO)
- 3 images missing alt text
- 1 image missing the alt attribute entirely
- 2 exact duplicate content pages
- 1 HTTP URL (should be HTTPS)
Forty broken links. Forty! That's a lot of 404s. And the worst part? Most of them were references to assets that didn't even exist—like assets/js/index.js and assets/stylesheets/index.css. Someone (probably me, months ago) had added these references to HTML files, but the actual files never existed. Classic.
Enter Cursor with Opus 4.5
Now, I could have manually gone through each file and fixed these issues one by one. But that would have taken hours, and honestly, I'm not that patient. Instead, I decided to see if Cursor with the Opus 4.5 agent could help me tackle this systematically.
I started by showing the agent the Screaming Frog reports. I had it read through the CSV files in my issues_reports directory, and then I asked it to fix the issues starting with the highest priority and highest number of URLs affected. The agent was smart about it—it created a todo list, prioritized everything, and then started working through the fixes methodically.
What Got Fixed
The agent went through and:
Removed Non-Existent Asset References
Found all the references to assets/js/index.js and assets/stylesheets/index.css that were causing 404s and removed them. Also cleaned up references to an assets/images/ folder that didn't exist. These were scattered across multiple blog posts and legal pages.
Fixed Broken Navigation Links
Updated relative navigation links that were pointing to non-existent pages. Changed things like href="about.html" to href="/blog/about.html" so they actually work. Fixed this across 10+ blog posts from different years.
Fixed the Double Path Issue
This was interesting. I had some blog pages that were using a <base href="blog/"> tag. Since those files were already in the /blog/ directory, this was causing URLs to resolve to /blog/blog/2019/deployment.html instead of /blog/2019/deployment.html. The agent pulled those files from the server, removed the base tag, and fixed all the navigation.
Added Missing H1 Tags
Added proper H1 tags to contact.html and legal/index.html. Changed an <h2> to <h1> in one case, and converted a <p>404</p> to <h1>404</h1> in another.
Added Alt Text to Images
Added descriptive alt text to three screenshots in one of my blog posts about deconstructing AI prompts. Also added an alt attribute to the US flag image in my footer that was missing one.
The Process
What impressed me most was how the agent worked. It didn't just make random changes—it:
- Read the issue reports to understand what was broken
- Created a prioritized todo list
- Read the actual HTML files to see the problems
- Made targeted fixes
- Updated the todo list as it went
- Committed everything with descriptive commit messages
- Pushed the changes to production
And it did all of this in one session. I just told it what I wanted, and it figured out the rest. The agent even pulled missing files from the server using curl when it discovered that blog/index.html, blog/post.html, and blog/contact.html existed on the live site but weren't in my local repository.
What I Learned
A few takeaways from this experience:
1. Regular SEO audits are important. I probably should have run Screaming Frog months ago. These issues had been accumulating, and I had no idea.
2. AI coding assistants are getting really good at systematic refactoring. The Opus 4.5 agent didn't just fix one thing—it understood the entire problem space, prioritized the work, and executed a comprehensive fix across multiple files.
3. The base tag is tricky. I'd forgotten I even had that <base href="blog/"> tag in some files. It seemed like a good idea at the time, but it was causing more problems than it solved.
4. Broken asset references accumulate over time. When you're copying HTML templates or moving files around, it's easy to leave references to files that don't exist. A crawler will find them all, but you might not notice until someone reports a 404.
The Results
After all the fixes were deployed, I went from:
- 40 internal 4xx errors → 0 (for the issues we could fix locally)
- 7 pages missing H1 tags → 2 fixed (5 are on a different subdomain)
- 3 images missing alt text → 0
- 1 image missing alt attribute → 0
Not bad for a Friday afternoon's work. The agent even updated my sitemap and committed everything with proper git messages. All I had to do was review the changes and approve the commits.
Now, I should probably run Screaming Frog again in a few weeks to make sure nothing new has broken. But for now, my site is in much better shape SEO-wise. And I have a new appreciation for what AI coding assistants can do when you give them a clear problem to solve and the right tools to work with.
If you're dealing with similar SEO issues on your site, I'd definitely recommend running Screaming Frog SEO Spider first to get a comprehensive view of what's broken. Then, if you have access to a good AI coding assistant, let it handle the grunt work of fixing everything. It's way faster than doing it manually, and you're less likely to miss something.
Plus, you get to watch the agent work through your codebase like a methodical code surgeon. It's oddly satisfying.
Enjoyed this post?
Get notified when I publish something new. No spam, unsubscribe anytime.