Large Language Models (LLMs) are no doubt impressive, but they’re not quite yet at Skynet levels. Their capabilities, though expanding, are currently limited when it comes to working with complex data sets and SEO data analysis.
Still, for SEO data analysis, LLMs like ChatGPT can be incredibly handy tools if you know how to make the data more easily digestible for them. I’ve seen my fair share of errors when working with vast numerical data sets, with both LLMs and humans. That’s why you always have to review the data!
I’ve learned a few tips through trial and error that I’ll share with you below. For a comprehensive understanding of AI’s broader impact on SEO and how to integrate these technologies effectively, check out our complete guide: AI in SEO: Essential Guide and ChatGPT, LLM Strategies.
Table of Contents
- Gathering and Tailoring SEO Data for LLMs
- Creating Prompts Optimized for SEO Data Analysis
- Analyzing ChatGPT Output for Accurate SEO Data Analysis
- Data Analysis Applications for ChatGPT and LLMs
- Conclusion
Gathering and Tailoring SEO Data for LLMs
When working with LLMs, it’s a necessity to clean and scrub the data you want to be analyzed. In my experience, ChatGPT and the other LLMs confuse metrics when they are all in decimal format. As many of our favorite platforms like to give us data in decimals to the nth degree, we have to take a few seconds to clean it up. Let’s take a look at how to do this.
Decimals – When you’re dealing with data you’ve downloaded from platforms like Google Analytics or Google Search Console, you’ll notice they’re often given up to a gazillion decimal places. The truth is, most of the time, you don’t need that level of precision. Neither your clients nor your boss are interested in endless decimals. Just keep it simple and round those numbers up. Stick to a maximum of two decimal places.
Percentages – Follow the same protocol here. Transform those decimal percentages into regular percentage terms and limit them to two decimal places if the decimal is necessary.
Monetary – Convert decimals into dollar values. This makes the data easier for the LLM to comprehend and work with. I recommend using full dollar amounts. Nobody cares about pennies.
Columns and Rows – Remember, your LLM, depending on which one you’re using, has a character limit. Plus, it’s best not to bombard it with too much information at once. So, limit the number of columns to 5 and the number of rows to approximately 25. By focusing on the top results, you can get the most efficient analysis from the LLM.
Let’s look at an example. Here’s an example of raw data:
Once we apply the best practices from above, it will look like this:
For this example, I ended up rounding everything in this example to whole numbers. The decimals were not needed to obtain a meaningful analysis of the data. Overall, it’s much cleaner using fewer characters and easier for LLMs to work with.
Creating Prompts Optimized for SEO Data Analysis
With our data prepped, now we can create an optimized prompt for your preferred LLM. I will be using ChatGPT for the following examples. For a deep dive into LLM prompts, be sure to check out the previous blog, SEO Prompts: Unlocking LLMS’ Full Potential.
For the initial prompt, we need to create a role description and objective, provide contextual information, the task to accomplish, any constraints, and the output format.
Role description – You are an SEO Analyst responsible for optimizing the website’s search engine performance and increasing organic traffic and revenue.
Objective – Your objective is to analyze the data from the monthly traffic and clicks report to identify trends, patterns, and opportunities for improving the website’s SEO performance.
Contextual Information – The provided data consists of 12 months’ worth of metrics, including clicks, impressions, click-through rate (CTR), average position, and revenue. The data represents the website’s performance in search engine results pages (SERPs) over a year.
Task – Your task is to perform a comprehensive SEO analysis based on the data and derive actionable insights to enhance the website’s visibility, increase organic traffic, and maximize revenue.
Constraints:
- The analysis should focus on understanding the relationship between clicks, impressions, CTR, average position, and revenue.
- Consider external factors such as seasonality or industry trends that may influence the website’s performance.
Output Form – Provide a detailed SEO analysis report that includes:
- Overview of the website’s performance trends and key metrics.
- Insights on the correlation between clicks, impressions, CTR, average position, and revenue.
- Identification of high-performing and underperforming months.
- Recommendations for improving organic traffic, CTR, and average position.
- Strategies for optimizing revenue generation through SEO techniques.
You can include the table data with contextual information or after the output form.
Analyzing ChatGPT Output for Accurate SEO Data Analysis
I strongly advise you to review the output generated by LLMs to ensure accuracy and reliability. In this section, we review the ChatGPT output for the SEO data analysis prompt from above, examining its insights on website performance, correlations between key metrics, and recommendations for enhancing organic traffic, CTR, and average position.
ChatGPT provided us with a numbered list that addresses each request from the original prompt. The Overview of the website’s performance and trends and key metrics is included below:
All in all, the analysis seems to be pretty spot on. The click rates look promising, and it correctly identifies the overarching trends for the year, among other things.
Let’s now peek at the second request, Insights on the correlation between clicks, impressions, CTR, average position, and revenue:
ChatGPT provided some correlation insights. but the big question is – how reliable are they? To cross-verify, I rolled up my sleeves and ran a correlation analysis on the data sheet, and here’s what it unveiled:
The data indicates a strong positive correlation (as we approach +1) between clicks and impressions, revenue and clicks, revenue and impressions, and revenue and CTR. In layman’s terms, as one increases, so does the other, and vice versa.
Meanwhile, between CTR and average position, as well as revenue and average position, we saw a pronounced negative correlation (as the data veers toward -1). This implies that when one data point escalates, the other one takes a plunge, and vice versa. Concerning the average position, this is quite logical. As the average position betters (say from 3 to 2), there’s a consequent boost in revenue, and the same goes for the click-through rate.
However, ChatGPT seemed to mess up the type of correlation for these two pairs. Despite the net effect being positive, it isn’t technically a positive correlation. In this context, it’s the negative correlation that’s triggering the positive outcome.
This mistake highlights just why it’s so vital to scrutinize any output generated by LLMs.
Let’s move on to the third ask, Identification of high-performing and underperforming months.
This looks good, let’s move on to Recommendations for improving organic traffic, CTR, and average position. What are ChatGPT’s recommendations? What’s ChatGPT’s take on this?
ChatGPT’s suggestions are pretty broad and general, but they seem to be headed in the right direction.
Finally, let’s delve into ChatGPT’s Strategies for optimizing revenue generation through SEO techniques.
Much like the earlier recommendations, these strategies are somewhat generic. However, they are pointing us in the right direction.
While ChatGPT provides a useful starting point for SEO data analysis, it is evident that careful cross-verification and human expertise are essential for ensuring the accuracy and reliability of the insights and recommendations.
Data Analysis Applications for ChatGPT and LLMs
There are countless ways to leverage LLMs for data analysis. Some applications I have used it for include:
- Keyword Analysis
- Content Performance Analysis
- Landing Page Performance
- Google Search Console Query Reports
- Google Search Console Landing Page Reports
- Goal Performance
Let’s look at how we can leverage LLMs for keyword performance analysis.
Keyword Performance Analysis
In the following keyword data, I have kept it short for space’s sake. You can use up to rows in ChatGPT, give or take. Here’s the keyword data we will analyze:
We want to find out any high- or low-performing keywords as well as any trends we should be aware of. The prompt format you can use:
Role Description: You are an SEO Analyst responsible for conducting data analysis to optimize the search engine performance of a dog-walking business’s website.
Objective: Your objective is to analyze the provided keyword data for the dog-walking business and generate insights to enhance its SEO performance. The analysis should focus on identifying keyword opportunities, understanding the relationship between metrics, and providing recommendations for improving organic traffic, CTR, average position, and revenue.
Context: You have been provided with a keyword report below that includes search volume, clicks, impressions, CTR, average position, and revenue data for various keywords related to dog walking services. The report covers one time period and includes percentage changes for each metric compared to the previous period.
Task: Your task is to analyze the keyword data, identify high-performing and underperforming keywords, uncover correlations between metrics, and provide actionable recommendations to enhance the website’s visibility, increase organic traffic, improve CTR, and maximize revenue generation.
Constraints: The analysis should solely focus on the provided keyword data for the dog walking business. External factors such as seasonality or industry trends are not included in the given data.
Output Form: The output should consist of a comprehensive SEO analysis report that includes an overview of performance trends, insights on correlations between metrics, identification of high-performing and underperforming keywords, and actionable recommendations for optimizing organic traffic, CTR, average position, and revenue.
Here’s some of the output from ChatGPT:
The output is detailed but will still require human review for accuracy. For more on ChatGPT and keywords, learn how to Elevate Your Keyword Research with ChatGPT and Competitor Keyword Research with ChatGPT.
Conclusion
ChatGPT data analysis can be a valuable asset for SEOs, with some limitations, of course. Their capabilities are expanding, but they currently have limitations in handling extensive and complex data sets.
However, by making the data more easily digestible for LLMs, we can harness their power as handy tools for SEO analysis. It is essential to note that errors can occur with both LLMs and humans when working with vast numerical data sets. Therefore, a thorough data review is necessary.
