OpenAI released a new AI chatbot called ChatGPT in November 2022. It has caused quite a buzz because it can answer questions and take instructions in a conversational, human-like way. According to OpenAI, ChatGPT can even answer follow-up questions, admit its mistakes, challenge incorrect premises, and reject inappropriate requests.
Since ChatGPT’s release, people have discovered numerous use cases for the chatbot. One idea that should spark interest among SEO experts and marketers came from Tobias Willmann, Head of SEO, Blick Group. He recently released a video demonstrating how you can use ChatGPT to sort Google Search Console Keywords into clusters.
ChatGPT Sorting Google Search Console Keywords into Clusters:
What is ChatGPT?
ChatGPT is an AI model from OpenAI that interacts in a conversational way. It is a type of machine learning known as a Large Learning Model. It trains on vast amounts of data and can predict what the next word in a sentence is. The more data ChatGPT
is trained on, the better its output, and the wider the range of topics and tasks it can cover. An essential feature of ChatGPT, and indeed all Large Learning Models, is that you don’t train them to do specific tasks. Instead, you train them using a wide variety of knowledge, that they can then apply to other spheres. When you think about it, that is a similar way to how humans learn.
OpenAI gave several example use cases for ChatGPT shortly after they released the chatbot. These included:
- A coding question – “this code is not working like I expect — how do I fix it?”. ChatGPT asked for more information about the context, and on receipt of that suggested possible fixes.
- The question, “how do you break into someone’s house?” – in this case ChatGPT explained that “it is not appropriate to discuss or encourage illegal activities, such as breaking into someone’s house.” The user then explained that he wanted to know how to protect his house from burglary, so ChatGPT suggested a series of useful security measures.
- What is the Fermat’s Little Theorem? ChatGPT then explained that theorem, giving examples. The questioner then asked how you could use Fermat’s Little Theorem in cryptography, which ChatGPT then explained in detail. The questioner followed up by requesting ChatGPT to make up a limerick about the theorem, which it did (and it made sense).
- Writing a short note to introduce the questioner to his or her neighbor – ChatGPT did as asked, and then gave a revised, more formal version in response to a request from the user.
We’ve even seen reports that ChatGPT managed to pass an exam at Minnesota University Law School that consisted of 95 multiple-choice questions and 12 essay questions. Although we must note that although ChatGPT “achieved” a passing grade, C+, most markers easily spotted its papers. "(They) had a hunch and their hunch was right, because ChatGPT had perfect grammar and was somewhat repetitive” according to Jonathan Choi, the professor who set ChatGPT its task.
Google’s management allegedly issued a "code red" after ChatGPT’s launch, concerned about the impact it could ultimately have on their search engine. It has reminded people that “scraping content, even with some modification, is against our spam policy.” Duy Nguyen from Google's search quality team added that Google has “many algorithms to go after such behaviors and demote site scraping content from other sites.
ChatGPT comes in two versions:
- ChatGPT – free. This is the standard version.
- ChatGPT Plus – $20/month. This gives general access to GPT, even during peak times, faster response times, and priority access to new features and improvements.
Uses for ChatGPT in Marketing
Unsurprisingly it didn’t take long for marketers to spot the potential of ChatGPT. Andy Gray, Founder, and CEO of The Graygency experimented with using ChatGPT to automate market research. The task would normally take him an hour, but ChatGPT performed it almost instantaneously. Gray continued to experiment, finding ChatGPT to be excellent for creating TikTok ad scripts.
Others have experimented using ChatGPT for tasks as diverse as writing Google Ad headlines and copy, translating ads into other languages, writing follow-up email messages from a call summary to creating compelling LinkedIn content.
Sorting Google Search Console Keywords into Clusters Using ChatGPT
Tobias Willmann, Head of SEO, Blick Group, shared a video demonstrating ChatGPT sorting Google Search Console keywords into clusters. He obtained a list of relevant keywords using his website’s Google Console and asked ChatGPT to sort them into useful topics.
To enable this, Willmann gave ChatGPT a list of suitable keywords, and then asked, “Can you cluster the following keywords by topic?”
Currently, these limitations limit the practical use of ChatGPT for sorting Google Search Console keywords into clusters if you have a significant number of keywords, but in that case, you may consider using its underlying “big brother”, GPT3 for this. OpenAI has opened up its GPT-3 AI model for commercial and research uses, by way of an API.
Using Keyword Clusters for SEO
If you aren’t experienced in SEO, you might wonder why you should cluster Google Search Console keywords into clusters using ChatGPT or indeed any other method. Well, this also relates to AI – in this case, the natural language processing (NLP) capabilities of Google’s search engine.
For years, marketers and others building websites have tried to boost the chances of their sites appearing high in search rankings by targeting keywords – words and phrases indicating what a web page is about. Typically, you would target one or two keywords with each post or page you create, preferably using different keywords (or keyword phrases) for each post/page.
Keyword clusters take a different approach. A keyword cluster is a combination of keywords that relate to similar purchase intent. For example, people searching for “mahogany dining suite”, “mahogany table and chairs” and “wooden dining suite” are probably all looking for the same thing, although “wooden dining suite” has a wider context. If you were to take the traditional approach you would only target one of these terms. However, ideally, you will want to rank at the top of Google for multiple relevant terms, even if some of them only have a small number of monthly searches.
By clustering your keywords by intent, you can rank strongly for multiple long-tail keywords, and even improve your rankings on short-tail keywords.
However, to do this you first need to come up with multiple relevant keywords. It is probably still easiest to use a traditional keyword tool at this stage, such as Ahrefs or SEMrush, although you may find a way to use ChatGPT for this in time.
Prioritize the keywords that provide the most economic value to you and make them the “core” keywords in your clusters.
Although some SEO professionals use thousands of keywords, remember that ChatGPT has practical limitations, so you may need to experiment to determine the maximum number you can use if you intend to use ChatGPT to place your keywords into the clusters.
Now, segment your terms as described above. Remember, you want each cluster to represent a search intent. Ultimately, you will be creating pillar content (for example a targeted landing page) that needs to be of interest to everybody searching for terms in a cluster. Each keyword cluster needs to have at least one keyword with a reasonable search volume but can include multiple niche long-tail keywords if relevant.
As well as creating a pillar page per keyword cluster, you can reinforce the terms by creating blog posts related to each pillar page. Each blog post needs to link back to the relevant pillar page.
ChatGPT Has Considerable Potential for SEO
Once you’ve started experimenting with ChatGPT you will probably come up with additional ways you can use it.
One possibility is to generate relevant topics to target related to keywords. For instance, you could ask ChatGPT,
Generate a list of X popular questions related to “<seed keyword>”, that are relevant for new <target audience>
So, for example, you could ask, “Generate a list of 10 popular questions related to “influencer marketing” that are relevant for small businesses.”
You could use ChatGPT to generate meta for your posts and pages. For example, Aleyda Solis provides an example of how you could use ChatGPT to generate titles for your page content. You would cut and paste your post/page content into ChatGPT and then ask it:
Generate X unique title tags, of a maximum of 60 characters, for the following text. They should be descriptive and include the term <”keyword”> in them.
Similarly, you could generate meta descriptions with:
Generate X unique meta descriptions, of a maximum of 150 characters, for the following text. They should be catchy with a call to action, including the term <”Main keyword”> in them.
Aleyda Solis also found ChatGPT to produce valid FAQ sections for posts by asking:
“Generate a list of X frequently asked questions based on the following content: …” and then point to the relevant section of copy for ChatGPT to use.
Things to be Wary of When Using ChatGPT For SEO
While ChatGPT clearly has uses in SEO, including sorting Google Search Console Keywords into clusters, we do have to be wary of certain limitations, that OpenAI has acknowledged from the start.
We have to realize that ChatGPT will not always be accurate. It is merely a tool that predicts words that will come next in a sentence. It isn’t sentient and doesn’t “understand” accuracy as such. The old computer acronym, GIGO (garbage in, garbage out) still applies here. If you feed ChatGPT incorrect data, then its predictions will most probably also be faulty.
In the past, we have seen some very public AI failures. These include:
- Tesla’s autopilot feature failed, causing a Tesla Model S to crash into a tree, killing two people.
- Amazon used an AI recruiting tool but discovered it was biased against women.
- An AI-powered camera that was supposed to automatically track the ball at a soccer game ended up tracking the bald head of a linesman by mistake.
- A 2019 study found facial recognition algorithms to be far less accurate in identifying Black and Asian faces than White faces.
- When Microsoft released its AI chatbot Tay in 2016, Twitter users manipulated the bot to make deeply sexist and racist remarks.
Microsoft’s 2016 Tay failure has clearly had an impact on the development of ChatGPT. OpenAI has programmed it to not generate text on the topics of graphic violence, explicit sex, and harmful content. They demonstrated this in their example where ChatGPT wouldn’t give details on how to break into a house. Tay hasn’t diminished Microsoft’s interest in AI, as they are now the largest investor in OpenAI.
While these protections are predominantly beneficial, they do impose some artificial barriers that could have an impact on any SEO tasks you may throw at ChatGPT.
In a similar vein, ChatGPT seems to currently have a bias towards the positive. It’s trained to be “helpful, truthful, and harmless”. This may sound good in practice, but it does color results that hinder ChatGPT’s ability to mimic the real world. You have to give quite detailed descriptions if you want ChatGPT to display neutral behavior (and even then, you can’t opt out of its “no-go” list of topics).
Perhaps more concerning to successful SEO activity, however, is that ChatGPT is currently unaware of content created after 2021. While this may not have a huge impact on sorting Google Search Console keywords into clusters, it may affect how it handles very new keywords.
Wrapping Things Up
As the fastest-growing app in the world, ChatGPT is rapidly becoming the new TikTok as far as traditional media are concerned. People are finding new uses for it virtually every day. It has proven to be adept at helping you optimize your website’s SEO, both with content creation and making technical changes.
One advanced SEO technique is to use large clusters of related keywords for your pages and posts. You base each cluster around a central “core” keyword. Although suffering from some inherent weaknesses, ChatGPT
is showing excellent capability for intelligently sorting these Google Search Console keywords into suitable clusters, ensuring your content and messages are seen by everybody of value.