Is AI the Evil Genie in the Lamp? - Lessons in AI Communication
Managing SEO for a multimarket site is never easy, especially when dealing with over 30 markets, each with its own peculiarities. Recently, my team faced a sitemap issue in our Drupal setup that led to several problems, particularly in multilingual markets:
- Unsecured URLs: Some links used
httpinstead ofhttps. - Incorrect domains: URLs pointed to the host provider’s domain instead of the correct one.
- Wrong language codes: Sitemaps mixed up country codes, like using
/mx/for Bolivia instead of/bo/.
Once we thought the issue was fixed, the next step was testing. But with no QA resources available, the idea of manually checking all sitemaps—market by market—came up. My response? Why do it manually when AI can help?
AI: The Evil Genie in the Lamp 🧞♂️
This is where things got interesting. I decided to to to ChatGPT to help me with the task, and it worked great. But working with AI often feels like making a wish to an evil genie. If you don’t phrase your request just right, you’ll get exactly what you asked for… but probably not what you actually needed.
AI shines with precision. The better you define your needs, the better the result. If your request is missing details, the AI will fill in the gaps with its own logic and get unexpected or quite frustrating results. 🤯
I found myself iterating again and again—refining prompts, adding specific conditions, and accounting for every edge case. But once I nailed it (a little less than 30m), the AI-driven solution analyzed sitemaps across multiple markets in minutes, pinpointing issues that would have taken hours to find manually. 🚀
📝 Tip: I personally love using the voice recording (dictation) option in ChatGPT because I can quickly give a lot of context to the AI.
The Takeaway
The real lesson here? Knowing how to communicate effectively with AI is an essential skill. Whether you’re automating processes, optimizing workflows, or simply trying to get reliable results—precision in defining your needs is crucial.
So, the next time you’re facing a tedious task, take a step back and ask yourself: Am I asking the right questions? If not, AI might just grant your wish… with a few surprises you didn’t see coming.
Cracking the Code with AI 💻
The process involved using a simple yet effective approach:
- Extracting and analyzing URLs from the sitemap.
- Checking for protocol inconsistencies (
httpinstead ofhttps). - Verifying domain accuracy based on the market.
- Ensuring correct language codes in URLs.
- Checking how percentaje of URLS has a numeric path (detect errors on alias generation)
Once I defined these checks clearly and structured my AI prompts correctly, I had a tool that could scan dozens of pages and produce results in no time.
Check out the solution
Here’s a CodePen demo where you can analyze sitemaps across multiple markets:
See the Pen XML Sitemap Validator by juandelaoliva (@juandelaoliva) on CodePen.
📅 Published on January 21, 2025
