Quick CSEs - a guide to making CSEs efficient for temporary usages (with AI)
Apr 12, 2025
Written By Kirby Plessas
Google Custom Search Engines (CSEs) are an under-utilized resource. They can be made quickly and robustly and shared with the whole team. And they don’t have to be permanent (or semi-permanent) tools. They are easy enough to make, with the help of AI, to create for one time uses.
Let’s first tallk about making CSEs quickly but without AI. My go to for this is to use Instant Data Scraper for Chrome. Any page that has a list of links I would like in my CSE is easily converted into a spreadsheet that I can use to copy and paste content into my CSE. Let me show you an example:
Similarweb lists websites and ranks them by popularity
Using a data scraper (consider using AI OCR to generate a list from a screenshot), I grabbed the top 45 online marketplace domains worldwide. Next, I used the Google programmable search page to create a CSE. I named it and put a dummy website as a placeholder.
I created the search engine, but next I chose to customize.
First update: I removed the dummy placeholder.
Then, clicking Add, I was able to paste in the contents from my source. I remembered to remove the top line (column name), and I added a few online marketplaces I knew were good but were not in the list.
The online link for the search engine is in the top panel of that page. I clicked through to do a couple sample searches.
Now that this search engine works, I added it to our Resource page.
The data scraper I used put things in a nice spreadsheet, making copy-paste into a CSE dead simple. But some lists aren’t scrapable this way, or, if they are, they need some cleanup before pasting into the CSE. Removing bullets, extra characters, descriptions, etc., is required for Google CSE, and the input field is very specific.
This is where AI comes in. Among the popular AI tools available, I found ChatGPT to be the best at understanding what I wanted, though any LLM should work. Then it is easy to copy a long list of websites with descriptions and bullet points, easily remove duplicates, clean up the text, and make it easily ready to insert into a CSE. For example, I was able to copy the results page from a Google search into ChatGPT and ask it to list URLs for the results only, without bullets, and it was ready to copy and paste into my CSE. I can even ask the LLM for a specific list of resources and add them directly into a CSE. Additionally, I could upload a spreadsheet or graphic to ChatGPT, have it resolve it, and then list the URLs for yet another CSE.
