Using ChatGPT Code Interpreter For SEO Analysis

So even though this has nothing to do with CustomGPT, I am having too much fun with Code Interpreter this weekend – so I thought about this simple use case for analyzing SEO data. 

The goal of this exercise was to start with data from ahrefs on some keyword research and then get a content plan of 10 blog posts to write based on that keyword research. 

So without further ado, let’s get started. 

How To Activate The ChatGPT Code Interpreter? 

Before we start any ninja analysis, let’s see how to activate the Code Interpreter. 

Please note: You need to be on the ChatGPT Paid Plan to get this. 

Step 1 : Click on the 3 dots (…)

Step 2: Enable the “Code Interpreter” in the “Settings”

Step 3: Activate the Code Interpreter Plugin

Click on the “GPT-4” tab and activate the plugin. 

ALERT: If this does not work, you might need to logout and re-login and do a Hard refresh of the browser (Control-R)

Step 4 : Code Interpreter Activated

If all is good, a Plus sign “+”  icon should appear in your inbox box. This is where you would upload your data. 

How To Get SEO Data For Analysis? 

For the purpose of this exercise, I am using ahrefs to get my keyword research data. I am using their “Keyword Explorer” tool to do some keyword research. 

Using this tool is outside the scope of this blog post, but there are many articles on that and you can also use their fantastic documentation and blog. 

Step 1: Conduct your keyword research and Export

For my use case, I started with some list of keywords of interest to my company and then click on “Export”

Step 2: Make sure you use “Excel” export

While Code Interpreter seems to be able to handle many formats, I saw that using Excel had best results. 

How to Analyze Excel Data In Code Interpreter? 

Now that we have Code Interpreter enabled and have our data, let’s get started.

Step 1 : Upload the data using the “+” icon

Notice how the Code Interpreter loads the data and even auto-corrects some data cleanliness issues. 

For us data analysts, this is a huge time saver – since data cleanliness and dealing with data issues usually takes the most time. 

In my specific use case here, it automatically went through 3 rounds of auto-corrections before the data was loaded. 

Step 2 : Examine the summary 

Now that the data is loaded, code interpreter will provide a summary of the data. Take a look and confirm that it has successfully understood it. 

Step 3: Run your prompt

Now that the data is loaded, you can give it your awesome prompts. In my case, my prompt was pretty simple: 

Excellent -- now based on this data, act as a expert SEO analyst and do a full SEO analysis and let me know what blog posts I should write -- give me the 10 article ideas based on this analysis.

Step 4 : Re-prompt or Chain-Of-Thought

If you don’t like what it has given you, you can do your “Chain of Thought” prompting and revise the output as needed.

Step 5 : Review your output

And here is the final output as markdown. For this, I just did this: 

excellent - please give this to me as a markdown table and also add other columns like slug and 160-char meta description I will need to write my blog posts. 

Frequently Asked Questions

Is this a replacement for Excel? 

I would NOT recommend that. While code interpreter and ChatGPT is great at text analysis, it is NOT a replacement for good old numerical analysis. 

You are better off doing your secret sauce numerical analysis in Excel or Google Sheets and then using Code Interpreter for the parts that need text analysis. 

So what is this Code Interpreter good for ? 

Code interpreter is mind-numbingly amazing at working with text fields. Check out this analysis I did for a different use case where it was able to run advanced text classification techniques like thematic analysis or sentiment analysis or even synthesizing patterns. 

Do I need to copy-paste data into ChatGPT ? 

No .. use the “+” icon to upload your data. It can take files upto 100MB. 

Notice that this is different from copy-pasting into the ChatGPT box. 

But wait – ChatGPT only has a small context window. How can it analyze 100MB of data? 

Code interpreter is different from the traditional ChatGPT where it looks at a small context window of 8K or 32K tokens. 

In this case, your data (upto 100MB) is actually loaded and analyzed using python code. 

You can see it working : 

But wait – it says “Code Interpreter” – is it only for “code” ? 

No – it can analyze data too – as evidenced by this use case. Yeah – OpenAI probably needs better naming – but it can do data analysis just fine.