ChatGPT has Turned SEO Upside Down AGAIN with Custom GPTs
| November 29th 2023
Just over a year ago, ChatGPT took the world by storm, introducing artificial intelligence into the everyday lives of millions of people. Who would ever have known that it was going to transform so many everyday tasks, slicing down the time it takes to complete routine jobs and freeing up time to focus on more meaningful and creative assignments?
ChatGPT has become a valuable (and low cost) marketing and SEO assistant to many businesses over the past twelve months. But using ChatGPT for things like keyword research, content planning, meta tag generation and schema markup ideas can now be considered just small fry. Because OpenAI, the brains behind ChatGPT, has recently introduced a way for anyone to create their very own SEO AI tools.
They’re called GPTs, and anyone can build one, tailored to their own business objectives. And even better, it doesn’t call for a huge amount of time or effort.
As long as you have a ChatGPT Plus or Enterprise subscription, you’re good to get started. So let’s take a look at what marketers and SEOs can do with GPTs, how they work, how to build your own custom GPT, and best practice guidelines from the chatbot itself.
What are GPTs?
GPTs are customised versions of ChatGPT that users can create themselves. All you need to do is brief the OpenAI GPT builder on what you’d like to create, and it will take it from there.
Whilst ChatGPT Plus or Enterprise users can already customise ChatGPT responses using custom instructions, the GPT builder cranks things up a notch.
Not only can you build as many GPTs as you like so you can hone each one to take care of highly specific tasks, you can also upload knowledge source files. So, instead of copying and pasting text from a whole bundle of resources as you would with ChatGPT, you literally just upload your files, and the GPT builder takes care of the rest.
If you have concerns over privacy, which many large language model users do, you’ll be pleased to learn that Open AI’s privacy controls allow you to opt your account out of model training. This means it won’t be able to access any sensitive information, such as your uploaded knowledge files or your chat history, in order to train their own models.
Do be mindful of potential data breaches though when uploading knowledge files, as it’s very early days with this technology and the risks are very real.
What can custom GPTs be used for in SEO and marketing?
Custom GPTs have multiple potential uses within the realms of digital marketing and SEO. Your fine-tuned model could help you with:
Whilst there is no substitute for the human content creator (and ChatGPT will itself back us up on that), you can build a GPT that is trained to generate the likes of ecommerce product descriptions, social media posts, blog posts and how-to guides, amongst other things. And if you feed it the right information, it should be able to match any preferences on tone, style and length too.
Just be sure to review everything that’s generated to make sure it’s on-brand, factually correct, unbiased and up to date. Always remember that ChatGPT itself says that it can make mistakes and that users should always check important information.
You can train your custom GPT on examples of well-optimised content to make it adept at generating SEO-friendly text. You can also set parameters, such as character length, for example.
This means you could, for example, create a GPT to produce the likes of meta descriptions, header tags, alt tags for images, slugs and any other vital SEO elements like these.
The GPT can be fed examples of code that align with the coding styles and preferences you want it to understand. The GPT will then, based on your prompt and its understanding of your preferred coding styles, generate a code snippet. This could be an HTML segment for a web page, or a CSS styling pattern for example. So, if you’re working on a specific project with unique coding standards, the GPT can be trained to stick to those standards.
But why is this valuable for SEO?
There are various benefits of using a custom GPT for coding, including speed of generation, error reduction, consistency and adherence to best practice. Ensuring your website code is error free and conforms to best practice is hugely important for SEO.
ChatGPT does, however, issue a caveat when using SEO AI tools like this, saying:
“While the model can generate code snippets, human oversight is essential to review and refine the output. This ensures that the generated code meets specific project requirements and quality standards.”
So by all means use your GPT to save you time and ensure consistency and accuracy with coding, but do make sure the output aligns with your overall objectives and organisation standards.
Schema markup is a type of code that you can add to the HTML of your website to help search engines better understand what the content on your web pages is all about. It tells the search engines what the content is trying to communicate, and converts unstructured data into structured data.
Schema markup can enhance a website’s ability to rank well in the Search Engine Results Pages. It can also enhance the appearance of your web pages in the search results by creating ‘rich snippets’, which are normal Google search results with additional information displayed.
Common rich snippet types include reviews, how-tos, events and recipes. All of them have the power to increase click-through rates, because they drive more relevant traffic.
Not all types of schema markup will lead to a rich snippet, but certain types will, so it’s an SEO tactic well worth pursuing.
However, it can be quite a challenge to identify – and write – schema markup. Even just a single misplaced or missing character has the potential to invalidate your schema, which would render it ineffective. And this is why your own custom GPT schema markup generator can be a very valuable tool.
You can start by using ChatGPT to identify schema markup opportunities for your website. All you need to do is ask it to tell you the most appropriate types of schema for your site based on its content, then copy and paste in your website URL, and see what it comes up with.
Once you are clear on what types of schema markup your website would benefit from most, you can go on to build your own custom GPT to generate the necessary code.
How to build a custom GPT?
In your ChatGPT account, click EXPLORE in the sidebar, then CREATE A GPT.
Enter your instructions in the message box. The GPT builder will then make a few suggestions based on what you’ve entered, such as a name for your GPT and a few conversation starters. You don’t have to accept the suggestions. Feel free to add your own, or use them as working titles and modify them later on.
Next you’ll be prompted to enter specific instructions so you can fine tune your GPT.
There’s a handy preview panel you can use to test your creation and hone it to suit your preferences.
For example, if you want more succinct responses, you can ask it to shorten its answers, or even stick to a particular character length. Or if you want to be able to cross check facts, you can ask it to cite its sources. This is particularly important if you’re using your GPT as a content creator.
Once you’ve got the basics nailed down, you can go to town on really tailoring your GPT to suit your specific objectives by using its advanced settings.
The advanced settings allow you to amend the instructions generated by the GPT building, as well as add your own guidelines covering how your GPT should function.
This is also where you get to upload your knowledge files. It’s seriously clever stuff, as you can provide anything from tone of voice guidelines to client personas and company values, and from prime examples of well-written SEO content to coding snippets.
You can also decide which of the default capabilities you want your custom GPT to use. These include browsing the web, creating AI-generated images using DALL·E, and interpreting code.
Finally, if you want your GPT to connect with other apps, you can use the ADD ACTIONS function to tell it which third party APIs you’d like it to work with, and how.
Once you’re done customising and honing your GPT, simply save it and decide who you’d like to share it with. You can opt to only use it yourself, to share a link with others, or make it public. Your finished GPT will appear in your ChatGPT side panel.
Building your own SEO AI tools – must-read helpful advice
Building your own custom GPT is a powerful way to tailor artificial intelligence to suit your own specific needs and business objectives. But there are a fair few things you’re going to need to bear in mind during the construction process.
Clearly define the task you want your GPT to perform
To get the most out of your GPT, it’s hugely important to take time to clearly define the task you want it to perform.
Clear prompts have always been essential when using large language models like ChatGPT. But when building your own GPT, the importance is magnified. It’s a bit like teaching a child to ride a bike. If you don’t tell them to pedal and steer, they’ll never get anywhere.
Set clear goals for your GPT
Also be clear on your goals for your GPT. What exactly are you looking to achieve? For example, if you’re using it to write SEO-optimised product descriptions for your ecommerce website, you’ll probably have an overall objective of getting your products ranking well in the Search Engine Results Pages, as well as converting sales. And if you’re creating a schema markup generator, you’ll want it to successfully achieve rich snippets to help drive more relevant traffic to your website.
Whatever you want your GPT to ultimately achieve, be sure to tell it so it understands what you are looking to accomplish.
Fine tune your GPT
Test your GPT over and over again, and ask other people you trust to test it too. If it’s not quite behaving as you want it to, it might be worth refining the knowledge files you fed it so that it’s more task-specific and has extra background to go on.
For example, if you are creating a GPT to write SEO-optimised ecommerce product descriptions, you may want to upload examples of what you consider excellent examples, either proven converters from your own site, or from a competitor site.
The quality of your knowledge files will directly impact the success of your GPT in meeting your objectives. Always make sure they are representative of the tasks you want it to undertake.
Evaluate your GPT
Assess the performance of your custom GPT on a validation dataset. This will help you make sure that the GPT you’ve created is learning the task effectively.
In evaluating your GPT, here are some useful points to factor in:
The purpose of your GPT… How well is your GPT performing against the purpose you set for it? It is actually learning the desired task off its own back, or simply memorising the training data? Ideally, you want it to learn the task off its own back so that it can apply its knowledge to different situations. This is way more valuable and can be likened to, for example, learning and reciting your times tables at school by memory, versus being able to work out how to do multiplication and division well so that you can apply the knowledge to other mathematics problems.
Validation dataset… A validation dataset is a separate set of data that wasn’t used when you first set up your GPT. Think of it as a benchmark to assess your GPT’s ability to generalise to new, unseen examples rather than respond by rote to your training data. The data in your validation dataset should represent the task, but should not be wholly identical to the data you input when you were training the model.
Metrics… Select relevant evaluation metrics based on the specific task you’ve trained your GPT to do. For example, if you’ve built your GPT to write SEO-optimised product descriptions, you might use metrics such as accuracy, keyword density or tone of voice.
Adjustments… Use your evaluation to identify any areas where the GPT is underperforming. If it’s not meeting your objectives, you may need to adjust your instructions, guidelines or capabilities, or change or add new knowledge files.
Apply iterative refinement
Iterative refinement sounds like quite a technical term, but it basically just means making adjustments to your GPT and its training based on the insights gained during your evaluation. It’s basically aimed at improving performance over time through a process of fine tuning that fits with your specific GPT task.
It may involve, for example, making changes to the model by adding or removing layers, or by trying a different approach altogether. You should also monitor performance on an ongoing basis against the validation dataset during each refinement cycle.
Always remember the importance of ethics and the human touch
As we have always said ever since ChatGPT and SEO AI tools first hit the mainstream, the importance of human intervention cannot be overemphasised.
Whilst AI tools are highly effective when it comes to cutting down the time it takes to carry out routine tasks, there is no substitute for in-depth knowledge and human experience, creativity and insight.
What’s more, as it advertises itself, ChatGPT ‘can make mistakes’. It also says:
“ChatGPT can be a valuable tool for marketers and SEO professionals, streamlining content creation processes and providing creative insights. However, human oversight and editing are crucial to maintain quality and accuracy.”
It is also important to be mindful of ethical considerations when using your custom designed GPT for any form of content creation. Be sure to avoid bias, misinformation or any content that could potentially breach ethical standards.
Finally, you should always adhere to OpenAI’s guidelines and policies when creating and fine tuning GPTs. And that’s from the horse’s mouth itself.
So, can AI help with SEO?
Absolutely, yes it can. And it’s getting better at it all the time.
OpenAI’s custom GPTs offer a personalised and effective way of getting the most out of AI for SEO and marketing content creation and optimisation, as well as a raft of other SEO tactics. It empowers businesses to create their very own custom SEO AI tools that align with their specific goals.
But do exercise caution in terms of bias, misinformation and inaccurate responses, and never cut out the human touch. And bear in mind that creating your own GPT that will drive the results you are looking for will take a good amount of effort and commitment to get it right. It’s definitely a case of the more you put into it, the more you’ll get out of it.
Embracing artificial intelligence – but never losing the human touch
At Figment, we’ve embraced the use of SEO AI tools. They have the power to speed up routine tasks and give us more time to focus on attending our clients and providing that all-important face to face service which, in the age of artificial intelligence, has come to be valued more than ever. It also gives us more time to focus on personalised strategies that help us achieve our clients’ ultimate goals.
If you’re looking for a London SEO agency that never shies away from embracing the latest technologies, but always remains client-focused, talk to Figment. We’re committed to the success of our clients’ businesses, and would be delighted to speak to you about boosting your online visibility with a tailored SEO campaign.