3/15/23 – Editor’s note – ChatGPT is now powered by GPT-4 instead of GPT-3.5. GPT-4 is much more powerful with mind-blowing capabilities. This post was written using GPT-3.5.
ChatGPT is all the rage right now with good reason. The capabilities of this software powered by the GPT-3.5 model is awesome. Like everyone else, we’re intrigued by this new AI chatbot and have been experimenting with it to better learn its capabilities. Lately, I’ve been trying to understand if ChatGPT can help perform search engine optimization, or SEO. This post shows some of my experimentation with ChatGPT and SEO analysis.
I asked ChatGPT to give me 30 keywords related to meat loaf recipe. Not bad. It gave me exactly 20 relevant keyword ideas. All ideas were broad, with each being only two words.
Although the list of ideas is accurate and useful, the terms are very broad and probably challenging to rank for. I next wanted to see if I could turn the list of 20 broad terms into one of useful longtail keywords. Ones that I could realistically target and expect to rank for if I had a recipe website. To do this, I simply swapped out keyword ideas with longtail keyword ideas in my question to ChatGPT, simple enough. Below is the resulting list.
This is why we’re all so excited about ChatGPT and its potential. By just tweaking my question to include longtail it produced 20 very high quality longtail keyword ideas for meat load recipes. Reading through them, they are excellent ideas which I absolutely think I could rank for if I had a recipe website.
Now I have nice list of longtail keywords that I can target with content. But, how much content is already targeting these keywords? Is another recipe needed? is anyone searching these terms? Does anyone really want a meatloaf slow cooker recipe? Let’s try and answer these questions by asking ChatGPT.
I asked ChatGPT; How many monthly Google search results does the query How to cook a perfect meatloaf in a slow cooker get?
Unfortunately, ChatGPT responded with it doesn’t have access to Google search results data, including the number of existing search results. This is a bummer. The number of SERPs for a keyword is a great way to understand the amount of competition. I manually googled the term and there are 5.09 million results. Not terrible, but not great either. Ideally, this is the number ChatGPT would have returned. It did however point us the in the correct direction by telling us to use the Google Keyword Planner tool.
Now I asked how many people are searching for How to cook a perfect meatloaf in a slow cooker on Google? Slightly different question. Here I’m hoping to understand not the competition for viewers, but how many people are interested in this recipe.
ChatGPT again tells us it doesn’t have access to Google search data. It instead points us again to the Google Keyword Planner tool. It also mentions popular SEO tools such as Ahrefs, SEMRush, Moz. That’s all fine and good, but not useful if we’re looking to use ChatGPT exclusively for our SEO analysis.
This is one big drawback of SEO analysis with ChatGPT – it doesn’t have access to proprietary information from the search engines, like Google. They keyword ideas are great, but if I can’t easily get the search volume and how many people are searching for the keywords, we have to guess if the keyword has interest.
I asked ChatGPT to pull keywords from the webpage https://divascancook.com/easy-meatloaf-recipe-best-meatloaf/. My goal was to see if I could use it to analyze keywords on competitor webpages. My specific question to ChatGPT was Pull the keywords from webpage https://divascancook.com/easy-meatloaf-recipe-best-meatloaf/ and list them in order of most times used.
It did a great job. It gave me the top 20 keywords from the webpage in reverse order of most times used. See the list in the below screenshot.
Next, I tried a recently published webpage. The page is on our blog and located here https://blog.wordbot.io/ai-artificial-intelligence/gpt-4-launched-watch-the-developer-livestream/. I asked the exact same question as above. As expected and seen in the screenshot below, ChatGPT failed to complete the request because it wasn’t trained on data that included this page.
With the launch of GPT-4, which is currently in the Pro version of ChatGPT, you can teach the model and make it aware of post 2021-data. That may solve problems such as this one. We could instruct the model to add the new webpage to its knowledge base before answering.
For link analysis, I again used the recipe link from earlier. I asked ChatGPT to list how many external backlinks the webpage has. The response is seen below.
This response surprised me. Again, we run into the 2021 issue, but that was expected. What I didn’t expect was that ChatGPT would actually be able to give us an answer. It responded with backlinks and referring domains as reported by Ahrefs. This is a very interesting situation because ChatGPT is seemingly using privately owned, proprietary information from a commercial SEO company. This demonstrates what I believe to be a coming battle in the field of AI – training on and using proprietary data without paying the companies that provide it. For our purposes though, we can see that we can in fact get SEO-related backlink analysis data from ChatGPT. Depending on the age of the webpage, the accuracy may be bad or good.
I asked What is the Google Page Rank of webpage https://divascancook.com/easy-meatloaf-recipe-best-meatloaf/. ChatGPT gave me a correct response, letting me know that Page Rank is no longer updated or maintained by Google as of 2016.
Finishing up, I asked ChatGPT to compare two webpages in a competitive analysis. I started by asking ChatGPT Which webpage has more relevant content regarding making meatloaf, https://divascancook.com/easy-meatloaf-recipe-best-meatloaf/ or https://natashaskitchen.com/meatloaf-recipe/
Both pages were created pre-2021. ChatGPT compared the two pages, recommended which recipe page was best, and backed its assessment up with a detailed explanation. It noted the favored page had more step-by-step instructions, more substitution options, and more photos. This could prove useful for making content improvement recommendations for content marketers.
Next, I asked for a numerical score for the two webpages. One we could use in a competitive analysis tool like wordbot. It didn’t want to give us a concrete score not knowing the scoring criteria.
So I gave it some scoring criteria and tried again. The Q & A from ChatGPT is below. Not only did it score the two webpages numerically, it broke down the score into each criteria highlighted in the question.
ChatGPT 3.5 is already useful for SEO analysis depending on the age of the webpage(s). With ChatGPT 4 which was just released, it has even grander capabilities. Its capabilities are limited as an SEO tool until it begins training on current data, but it definitely still has value.