AI content has been a hot topic for about a year now.
Before ChatGPT entered the scene, people were already leveraging AI for content creation, but such content was challenging to rank.
It didn't read as well, and scaling it was difficult.
However, everything changed with the advent of ChatGPT.
It opened up numerous possibilities in content creation and, within just 12 months, profoundly impacted the content creation industry.
For individuals like me, who actively run automated link-selling websites, AI content has been a blessing in disguise.
That said, AI content isn't suitable for every situation. Knowing when and how to use this tool effectively makes a significant difference.
For instance, I haven't used AI assistance in any form for this website. It simply doesn't fit, and forcing it would yield poor results.
However, for many of my other web properties, especially those that are automated and sell links, AI content has been producing the desired outcomes (at least for now).
This is why I've developed the following AI content creation framework, which I'm eager to share with you:
- Custom GPT for brief creation (the "old ChatGPT" works just as well).
- Cuppa.sh for an actual AI content development. Cuppa allows you to use your own API key from OpenAI, significantly reducing costs.
- Custom GPT for internal linking.
- Custom GPT for proofreading and light editing.
But before we delve into the specifics of how this framework works (and what types of prompts you should use), let's review some results of AI content.
Site #1 - Tech Niche
Here's an example of a website where I've been leveraging AI content for the past 6 months:
The website is experiencing a notable hockey-stick growth pattern. There's a decent amount of clicks and impressions, and it's clearly trending positively.
This is encouraging, especially considering that Google has been penalizing many sites during its recent updates.
Unless your website is Reddit or Quora, you've likely seen a decrease in traffic for at least some of your keywords.
A few important things to mention:
- This website targets low-competition keywords in the tech niche.
- It is not a brand-new site; it has been around for nearly two years before I began leveraging AI content.
- The site is hosted on an aged domain.
- It possesses considerable authority, currently boasting a Domain Rating (DR) of 40.
- The site has a well-established SEO strategy.
Site #2 - Design Niche
Here's another website where AI content seems to be producing decent results:
However, unlike the earlier example, this site is not ranking nearly as well.
However, unlike the earlier example, this site is not ranking nearly as well.
The number of clicks is quite low, and while the AI content seems to be getting indexed and picked up by Google, it's not appearing on page 1 for a lot of its keywords.
A few important things to mention:
- This website targets low-competition keywords, but it also aims for some medium and high-competition keywords.
- The site is hosted on an aged domain with some authority.
- It has been used as a platform for selling link placements for nearly three years.
- The site lacks a strong SEO strategy, with both technical and on-page aspects being fairly weak.
Site #3 - Business Niche
Here's an example of a website where AI content made no significant difference:
Even before I started implementing AI content, this website wasn't showing much promise.
The AI content didn't seem to help, but considering its deficiencies in other areas, I can't solely attribute this to the AI.
Of the 200+ articles I've added, I didn't use the AI workflow I'll describe below for about 180 of them.
Initially, I attempted to upload 180 AI-generated articles in one quick batch (bulk mode), hoping for an immediate boost, but it was ineffective.
This site has a setup nearly identical to the previously mentioned website, but I also used it as a testing ground for some of my own experiments.
I should allocate more time to improve this website and see if that enhances its rankings.
What Did I Conclude About AI Content So Far?
Here are my top conclusions:
- AI content is only one piece of the puzzle. To succeed with AI content, you need other elements in place as well, such as solid SEO fundamentals and, ideally, some level of authority.
- Bulk content creation may be effective, but only for extremely large quantities (1,000+ articles, covering specific topical maps). However, it feels like a churn-and-burn strategy that is too easy to spot.
- There are many things I haven't yet leveraged AI for. In scenarios where AI content is applicable, it can easily rival human-written content. It comes at a significantly lower cost, requires less time, and is much easier to scale.
- As for ranking AI content, it doesn't seem to be much different from human-written content. To rank AI content, you need authority, links, and good SEO fundamentals – essentials you would need for any content.
Do I like using AI content?
At the moment, I am somewhat indifferent.
In an ideal "Internet" scenario, ChatGPT would never have been released, AI would never have become a thing, and we would write solely for users.
Google would know how to detect the best content for every query and wouldn't rank Forbes for everything.
But shit, this isn't the ideal scenario. There never was one, and we're unlikely to ever see it.
AI is now a part of our lives and will continue to be. The genie is out of the bottle, so there's no use in denying it.
And Google has been a dumpster fire ever since its creation. So, I'm going to continue doing what the algorithm asks of me – I'll produce content that seems to rank, and I'll do it for fun and profit!
My Simple AI Content Workflow
Now that I've covered my results and my "expert" AI analysis, let's dive into my workflow.
I find this workflow to be quite simple and straightforward - primarily because guys over at Cuppa.sh have built a nice tool that does most of the work for you.
And once you train your own GPTs, it's mostly plug and play.
ChatGPT For Brief Creation
Before I developed a custom GPT, I was operating a regular chat where I instructed ChatGPT to use WebPilot to scan Google's top 10 results and then create a custom content brief for a specific keyword I provided.
My prompt was essentially:
Using webpilot create the heading structure that is perfect for this keyword: [KEYWORD]
I only need headings, not additional comments or anything else. I would love to make it as granular as possible, so make sure we have h3s and h4s.
The results were mixed. Often, he would create something somewhat generic, so in those cases, I would provide additional prompts to achieve better results.
For example, I might say:
The outline seems generic and not as based around the main keyword. Put yourself in the shoes of a reader who is searching for this keyword. Make sure you provide specifics in the brief, and make sure that the brief answers all of the important questions
I tend to experiment with prompts until I get the desired result, but often, he gets it quite right from the start.
One prompt I particularly like is asking him to create additional FAQs based on the 'People Also Ask' section, as well as questions from other content he scanned.
However, I don't always do this, as for some straightforward topics, I find an FAQ section that just repeats the same information unnecessary.
Here's a result he gave me based on 1 prompt only for this keyword: how to make $200 a day with Instacart
Not too bad, and I'd be happy to run with this, to be honest.
Is this enough for top-level content creation?
Not always. If this was for a website where I need to ensure that the content is the absolute best it can be, I would go and further modify that brief. And how do I do that? Through manual research.
Fortunately, I don't need to do manual research because Cuppa allows me to view content from other sites in the SERPs.
Creating Briefs in Cuppa
When you input your keyword into Cuppa for content development, Cuppa analyzes the headings and content from competitor sites in the SERPs.
This is incredibly helpful because it allows you to instantly see what others have written about, enabling you to modify and add to your brief accordingly.
The user interface is also well-designed, facilitating quick addition and editing of headings directly from your competitors' lists for your own content.
Cuppa.sh - Content Development
Once your brief is ready, it's time to develop the content.
As shown in the screenshot above, Cuppa offers several interesting options. Like some other AI tools, it allows you to choose:
- Tone of voice (with several different options).
- Point of view.
- Generation of an FAQ section.
- Inclusion of YouTube suggestions.
- Generation of a meta description.
- Inclusion of a featured image (from Unsplash).
Personally, I always opt for an SEO-optimized or friendly tone of voice, depending on the content I am producing.
First-person point of view is also a standard from which I don't deviate.
Regarding the FAQ section, I don't rely on Cuppa for this; I manually add the FAQ section and each individual question, if I find any relevant ones (or if ChatGPT generated them when creating the brief).
Adding Research Data
One feature of Cuppa that I absolutely love is the ability to add "Research Data" for any heading I create in an article.
This is perfect for making the content less generic and ensuring that much of the manual work is handled by the tool.
Advanced Options
Cuppa enables the incorporation of additional prompts to create more appealing content.
I am still in the process of determining the most effective prompts.
Sometimes, certain prompts work exceptionally well, while in other instances, they may not be as successful.
It's all about experimenting and observing the results firsthand.
Another significant advantage of Cuppa is its community.
They maintain an active Discord channel where both owners and users share their suggestions.
This interaction allows you to see what others are experimenting with and gives you the opportunity to try these ideas yourself.
With an abundance of innovative ideas and prompts available, joining the Discord channel is highly recommended.
Brew Time
When the structure feels ready, I hit "Generate Article".
The time it takes for Cuppa to generate content can vary.
This waiting period depends on your prompts and heading structure, but primarily on OpenAI's current capacities.
Typically, the wait time for a single article ranges between 10 and 20 minutes.
Cuppa.sh vs Alternatives
It may seem like I am overly promotional of Cuppa, and to some extent, I am. I've been a fan of the tool from the beginning, appreciating their trajectory and the developments they've made this year.
Are there good alternatives? Absolutely.
Koala.sh is another tool that produces content of similar quality.
Most AI content writing tools don't offer this feature, too.
Among the tools that I've tried and worked with, only ZimmWriter allows you to integrate your own key.
I've heard great things about it, but I decided against using Zimm because it is a desktop-based Windows tool, and I preferred something online that allowed me to directly upload content to my WordPress sites.
Custom GPT for Internal Linking Purposes
Before the latest ChatGPT release, I was doing this the "old-fashioned" way of creating 1 dedicated chat where I was feeding information and data to ChatGPT.
Essentially, I would provide it with my website data, including sources for internal links, desired destinations for these links, and a variety of anchor text options.
I would then prompt ChatGPT with my internal linking criteria and instructions on what I wanted it to do.
The process that emerged was as follows:
- ChatGPT, using WebPilot, would analyze the content from my provided list of URLs.
- It would determine if there was an opportunity for internal link placement towards any of the target URLs.
- It would suggest what content needed to be added or rewritten to accommodate the new internal link with one of the anchor texts I had provided.
That's it.
With the latest GPT release, you can create your own GPT, and train him until he gets this completely correct, along with your desired internal linking criteria.
Custom GPT for Proofreading/Editing
Someone might ask me here to provide the actual GPTs I created and make them accessible to everyone.
While I could do this, I believe it would be a mistake.
Given how easy it is to create custom GPTs for specific use cases like proofreading and internal linking, there's little reason to rely on someone else's model.
This is particularly true because you should tailor your GPT to your own data and criteria.
If you're running a website in a specific niche, have a unique tone of voice, or incorporate any distinct elements, training your own GPT will yield the best results.
That's essentially what I did.
Additionally, I trained my GPT to replace certain words, phrases, and sentences that are most commonly used in AI content creation, making each content piece sound less generic.
Whenever Cuppa completes content creation, I simply pass that content through my GPT proofreader.
What I Don't Use AI For
While AI, in all its power, helps us automate and speed through a ton of manual work, I still don't see it replacing many processes.
Here's what I don't use AI for:
Keyword Research
When ChatGPT first came out, I saw a ton of people on Twitter asking it to do keyword research.
Any SEO worth his salt would tell you that's a bad move.
While AI might be good for brainstorming different topic ideas, an actual keyword research process needs to be backed up by reliable quality data. It also needs to be performed in a tool that gives you an overview of SERPs, along with many other metrics that are relevant in the KW research process (like backlink data).
For KW research purposes I leverage following tools:
Many people are shitting on Ahrefs because of the whole drama around the credits system (deservingly so), but I think it still remains to be an absolute best SEO tool.
Final Edits
Before I hit publish on content, I often go in and do a final round of editing.
I am trying to replace content that sounds too generic, or things that are obviously not factual based.
There's no way to automate this process because this process exists due to automation I've set up earlier.
Content Repurposing
I would love to discover an AI tool that effectively handles content repurposing, but so far, I haven't had any luck.
As a result, all the content I create for these websites that I want to share elsewhere needs to be repurposed manually.
While I enjoy this part of the process, it can sometimes be time-consuming.
Big Question - Can You Scale AI Content Creation?
You absolutely can (and probably should) scale content creation using AI.
To do this successfully, you will likely need to hire multiple people to manage aspects of the content creation process that are not suitable for AI.
It's also important to note that many people are scaling content creation using AI independently, leveraging various tools.
For a significant number of these individuals, this approach is proving quite successful. Is it a long-term strategy?
Probably not, but it can generate a substantial amount of cash very quickly.
Ultimately, how you choose to employ AI in your content creation process is entirely up to you.
My goal is simply to share the results and details of my experiments, hoping to inspire you and provide a perspective that may be helpful.