Table of contents:
- The power of ChatGPT for SEO
- Mastering prompts for AI bots
- How to overcome the “writer’s block”
- ChatGPT SEO clustering: grouping keywords at scale
- Getting initial SEO tips when beginning as a newbie in the field
- Extracting entities, their relationships and constructing knowledge graphs
- Simplifying writing regular expressions (regex natural language queries)
- Train the bot to generate similar paintings on a same topic like your previous art pieces
- Ok, then, will ChatGPT replace Google?
The Power Of ChatGPT For SEO
The Internet and social media are exploding with news about OpenAI’s new AI product, ChatGPT. It is a new generative language model optimized for chat or in simple terms you can send prompts to their AI Bot (requests in the form of mostly short texts in natural language) and get an answer back (a textual answer to a question, initial article drafts, summarized responses, code in various programming languages, you name it).
We already wrote about prompt engineering and generative AI for SEO in some of our past articles but it seems that the world did not get their full power before ChatGPT was introduced to the wider audience. We are proud to share that we have been some of the pioneers in this field and we actively worked on developing SEO friendly workflows for text generation in the last two years with specific implementations for eCommerce and news websites.
What we have learned is that large language models like this are trained on a large amount of data and require significant financial investment in the computer, so it is not easy to replicate them. Models like this are capable of performing tasks for which they have not been explicitly trained, such as translating text from one language to another [e.g., from English to German and vice versa] and applying that knowledge to other domains.
You can tell that it is worth giving a go if you consider the fact that ChatGPT got 1 million users in the first 5 days of launching, something that took Netflix to achieve in around 3.5 years, Facebook 10 months, Spotify 5 months and Instagram in 2.5 months! Crazy! Some SEO marketers, like Malte Landwehr, even predict that “a lot of typical outsourcing or junior jobs in management consulting, legal, finance, etc. will be replaced by AI over the next 2-10 years. In such roles, companies will fire 90% of the team, keep the top 10%, double their salaries, and 5x their efficiency with AI tools like ChatGPT”. Are you convinced now? Because the hype is there and it will be here to stay!
The real power that ChatGPT has in general is that it is able to give direct answers to questions through its simple user interface: we are able to get responses back in a matter of a few milliseconds, instead of the casual search process through pages and links, like we do on search engines like Google. As a consequence, ChatGPT caused “code red” at Google. Apparently, Google’s CEO Sundar Pichai participated in urgent meetings to further define the AI strategy for Google after the release of ChatGPT. Having in mind that their business model heavily relies on ads and e-commerce features that are the basis of their revenue model, sources reveal that Google finds this concerning.
Let us dive in more into details how you can use ChatGPT SEO for your content marketing strategy and as a productivity tool to develop your ideas faster. Fasten your seat and let’s go! 🙂
Mastering Prompts For AI Bots
First things first. The first step: SEOs needed and still need to master prompts. That is because prompt quality dramatically affects the quality of the given output. Those who manage to be successful in doing so, putting the effort to learn the basics about how an AI chatbot generally works and embrace new technologies, will become leading voices in the industry and dominate the future. You will need to write clearly, to the point and descriptively enough to explain your prompt request.
Two years ago it was really primarily about prompt mastering, however, nowadays LLMs have evolved and we know that a successful SEO shall take into account:
- the data that will make the prompts unique (this is why you have an extra competitive edge if you have invested on structured linked data);
- have a content validation pipeline (because LLMs are statistical machines with no real understanding of the content they produce);
- work to embed the tone of voice of the brand (you need transfer learning, using fine-tuning or in-context learning);
- be responsible when using AI and most importantly find the best feedback loop by joining forces with domain experts and content creators.
LLMs are helpful and worth the effort but they can be complex as well. One way to judge their “complexity” is their size. The number of their trainable parameters per LLM can vary – more parameters mean more training but at the same time more parameters can mean that the model is able to solve more complex problems. Another way to measure the size of the model is to consider the amount of data it has been trained on. Finally, the reason for enormous model sizes, no matter how they are measured, is the application of transformer architectures – they are the main driver behind the latest innovations in the NLP, NLG and NLU field. We even predicted LLMs’ advancement 2 years ago: now it is your turn to embrace their full power.
Try our AI question answering tool to extract relevant questions from URLs and then automatically answer!
How To Overcome The “Writer’s Block”
Engineering good prompts is not easy but it can be learned. Here are some example prompts:
- “Write me a summary of Harry Potter”
- “Write an article on the topic of knowledge graphs” (btw, this article is written by a proper human, no worries)
- “Showcase your previous answer in bullet points”
- “Explain the benefits of using ChatGPT in a few sentences”
As you can see, the verb is leading in the beginning, then we have the subject and finally a guideline about how we want to structure the answer. For an ideal prompt engineering, we need to follow the MCSSCF prompt framework:
- Write proper meta instructions which describe the expected output, e.g. “provide an answer to X as you were an expert computer scientist”;
- Provide the right content modifiers, e.g. “a summary of…”;
- Subject description, e.g. “black jacket for women from the 80ties”;
- Styling, e.g. “provide an answer in the style of Shakespeare”;
- Context, e.g. provide some text for context and then ask questions about it;
- Format, e.g. camera details, text styling, resolution output, etc;
- And so on, you can learn more about how to perfect your prompt engineering here.
To summarize, it is absolutely vital to be clear, concise and to the point!
If you are having trouble coming up with ideas for your writing, the suggestions above can help you overcome your writer’s block. In the first phase, start by writing some prompts to ChatGPT and then refine them. We do not expect the AI bot to answer absolutely perfectly, but it’s still much better than many other AI tools out there.
ChatGPT SEO Clustering: Grouping Keywords At Scale
Tired of paying for SEO tools to cluster your keywords at scale? Not having the right coding skills to do it by yourself or at least the program is not fast enough to do this properly fastly and reliably? Then ChatGPT comes to the rescue! You can give an order to the bot through the prompt “classify the keywords” and provide the list of keywords on your side and you can expect them to see them clustered based on their semantic closeness. Easy, fast, reliable: say bye bye to complex Excel sheets commands, expensive software or complicated AI tools that require a bit of a learning curve in order to learn how to use them properly! Tobias Willman had a wonderful Python script on the topic that you can analyze and try to adjust to your specific case.
Getting Initial SEO Tips When Beginning As A Newbie In The Field
Are you a new SEO trainee who just barely started out in the SEO world? There is nobody that can act as your mentor to point you to the right literature, especially when solving specific problems? No worries, ChatGPT SEO got you covered!
Ask “How to do SEO for an e-commerce website” or instruct the AI chatbot to write a text for you by combining multiple, connected prompts like:
- “Include H1, H2, H3 headings in the text”
- “Use specific entities in the text”
- “Include specific keywords through the writing”
and voila! You have your first ready content piece that you can feature on your blog. Easy peasy.
Extracting Entities, Their Relationships And Constructing The Basis for Knowledge Graphs
Are you an SEO professional who does not have any specific experience with the computer science and engineering world but still loves to play with the newest AI technologies whenever you get the chance to do so? Not versatile in the NLP world but already know the basics of entities, entity relationships and the benefits of having a content knowledge graph? You are in the right place! You can instruct ChatGPT to extract entities and entity triples to form a knowledge graph and its appropriate ontology. As Lee Foot showcased on Twitter, we can use the following prompts in order to make this possible:
“1. Build a knowledge graph in RDFlib format for…
- Build ontology in OWLready2 format for…
- Extract semantic triples in a dataframe with columns, s,p,o (Python code) for the following text…”
In any case, please note that ChatGPT is still unable to create entity links to demonstrate true semantic understanding, and sometimes the bot star can be misleading: it can create markup using schema properties that do not exist, and it is also able to create Wikipedia links to pages that never existed! Crazy, is not it?
This is definitely a shortcoming of ChatGPT. Be careful how you use this AI beast – wrong intentions, bad prompts and too high expectations can lead to bad or even harmful results. Be careful!
Simplifying Writing Regular Expressions (Regex Natural Language Queries)
Have you ever tried to find the right regex expression to detect or remove substrings in your text? We must admit that we are among those who struggle with writing proper SEO regexes and search for the right programming trick on Stackoverflow, just like you. Now you can solve this problem almost completely by telling ChatGPT to do it for you, all in simple, natural language!
Here is one regex instruction that Aleksa Gordic, research engineer at DeepMind, posted on Linkedin a while ago:
Train The Bot To Generate Similar Paintings On A Same Topic Like Your Previous Art Pieces
Boost your content designers and graphic designers with ChatGPT by training it on your own art pieces from the past. You do not need to go to Shutterstock and other stock image-based websites anymore to buy the right picture for you. You can start making a triple collaboration between you, your designers and ChatGPT and ask the AI Bot to “create a picture on the same topic like the one shown before” or “create a picture of Googlebot in the style of X”. It is all about prompt engineering.
Ok, Then, Will ChatGPT Replace Google?
The short answer is NO. Consumer search engines are processing billions of queries that require information extraction from sources that exist only a few hours, days or weeks. At the current moment of speaking, in order to achieve the same result, LLMs will need to have the ability to be retrained at the same speed interval and ChatGPT does not satisfy that criteria. For now.
You can justify this fact just by trying to write up some product queries and you will see that responses are mostly, if not all, containing information that it is dating up to 2021 but 2022 data is not included. This affects the freshness of the given responses and of course this compromises their truthfulness.
Long story short, we are still lacking exact, in-depth information on how LLMs are trained and how exactly their black box AI magic is working behind the scenes. At the same time, queries provided to search engines are longer and expressed in natural language, not really in a prompt-based fashion and search engines like Google are already winning and dominating that front.
The future belongs to those who embrace technology. You should not be afraid of it, but rather learn how it works, educate yourself on the topic of prompt engineering or collaborate with your developers to help you out on the way.
We need to be absolutely aware of the importance of having a strong data foundation and validation – without solid data foundation and knowledge graph strategy it will be impossible to achieve a highly-performant SEO organic strategy and be prepared for the Generative AI era. Just like Andrea Volpini predicted the SEO trends for 2023:
“Automation and augmentation remain closely intertwined, and the ability to experiment with new optimization techniques such as in-context learning, instructions-based prompts, and other emerging behaviors of foundational models remains anchored to an extensible data fabric and brand authority.”
AI and ChatGPT SEO are here to stay, the question is….will you be part of the revolution?
More Frequently Asked Questions
What is ChatGPT?
ChatGPT (or also known as Generative Pre-trained Transformer) is a chatbot AI product developed and published by OpenAI (a San Francisco company) by the end of 2022. It is built based on the GPT-3.5 family of large language models where computer scientists used both supervised and reinforcement learning techniques to fine-tune this AI model.
Is ChatGPT free?
ChatGPT was free in 2022 and still is during its research phase, however, it is expected that the tool will be monetized in future.
What can ChatGPT be used for?
ChatGPT can be used for many tasks and it can be used for a variety of language tasks, such as language translation, summarization, question answering, and text generation. For SEO purposes, some of them include generating content in the form of text, like clustering keywords, generating structured data and robots.txt files, writing code, extracting entities and their respective relationships, creating text summaries, creating social media posts or website content and so on.
Can ChatGPT translate language?
You can use ChatGPT for translation: simply enter the text that you want to translate into the chatbot and specify the languages that you want to translate between. For example, you might instruct the tool: “Translate ‘Hello, how are you?’ from English to German.” and it will generate the requested translation.
It’s important to keep in mind that machine translation is not always perfect, and the quality of the translation will depend on the complexity of the text and the languages involved. However, ChatGPT can provide a useful starting point for translation tasks and can be a helpful tool for understanding the general meaning of text in a different language.
Is GPT-3 the most advanced AI?
This model is certainly the most powerful and complex AI model, consisting of billions of parameters. However, it is not necessarily advanced in every possible sense. GPT-3 is particularly strong at natural language processing tasks, but there are other AI systems that are designed for tasks such as image recognition, speech recognition, or robotics, and these systems may be considered more advanced in those domains. It is hard to define which AI system is the most advanced at any given point in time: everything depends on the specific capabilities, intents and goals of the user.
Must Read Content
The Power of Product Knowledge Graph for E-commerce
Dive deep into the power of data for e-commerce
Why Do We Need Knowledge Graphs?
Learn what a knowledge graph brings to SEO with Teodora Petkova
Generative AI for SEO: An Overview
Use videos to increase traffic to your websites
SEO Automation in 2023
Improve the SEO of your website through Artificial Intelligence
Touch your SEO: Introducing Physical SEO
Connect a physical product to the ecosystem of data on the web