How to Build a Custom Content Writing GPT

(Disclaimer: Profanity.)

In this chapter, I’ve mentioned the realistic benefits of building your custom content writing GPT, certain principles and philosophies that guide that process, and the process itself.

This is a detailed guide due to the methodic approach. Feel free to bookmark this for later use.

Expected benefits of custom content writing GPTs

Creating a custom GPT for your tailored content needs is a multilayered and complex process that requires you to commit hours of deep work. It is crucial to start with the correct expectations as it will help you evaluate the ROI on your behalf.

These GPTs are not magic bullets or a better version of GPT-4o. 

They do contain some portion of your prompt so that your workflow becomes faster. 

For example, if you are writing LinkedIn posts, a part of your prompt would contain platform-specific instructions such as “write for a professional audience” or “include whitespace for readability.”

When these segments are already coded into your custom GPT, you can produce LinkedIn posts quickly. Fundamentally, the quality of output will remain the same.

For most of the use cases, individual tasks will be quicker, considering the reduced prompting needs. However, there could be instances where the opposite could be true. 

For instance, if you are writing an informal LinkedIn post, and your custom content writing GPT is coded to produce geeky stuff, you are fucked. In those scenarios, you have to provide extra instructions to overwrite the existing ones.

Regardless of whether these tailored AI bots save you time during one individual task, they do save you a lot of time in workflows. The quality of the output is substantially higher and the time saving is quite noticeable.

To give you an estimation, I take about 4 hours to write a well-researched 1200-word article. My ChatGPT content writing workflow, which contains a bunch of these custom GPTs for various purposes, helped me reduce it to 2 hours and 30 minutes!

Let’s learn why this approach works.

GPT-creation principles for content writing

The most important principle of custom GPT creation is to build them for simple tasks only.

Now, WTF are “simple tasks?”

Well, tasks that can be done with one-shot prompting. You just paste your problem statement and the produced output is what you are looking for. Whether you directly use it as is or refine it further based on your taste, you should get the desired result in one step.

The Meta Details GPT is a great example of this. All you have to do is paste the whole article, the outline, or just the title, and it will immediately provide the meta description, alternate title, and URL slug suggestions.

In fact, all of the GPTs in my ChatGPT content writing workflow linked above are built on this philosophy to maximize efficiency and performance.

As you can imagine, the task must follow a pattern to fit into this principle. You can easily do that by breaking your workflow into small chunks. Simple, patterned, one-step tasks perform exceptionally with GPTs. This includes research-oriented tasks too.

The second and last principle of creating a custom GPT for content production is related to the quality of the LLM’s answers. Of course, quality is subjective as it depends on a myriad of factors. 

Instead of boring you with them, let me tell you how my definition of a useful output:

1. If the content piece is around 50-80 words, it should require minimal editing, if at all.

2. Longer content pieces should contain enough information to be used as a reference while I write them myself.

To put it into perspective, for meta details, the output should require zero editing.

For article outlines, the output should contain enough data to serve as a solid resource for me.

Build a custom GPT for content writing

Break your prompts into three parts:

1. The algorithm: This is the operation you want the machine to do. Whether it is refining existing content or synthesizing new stuff.

2. The problem statement: Here you mention the details related to the task at hand.

3. The output description: It includes details like the tone, format, structure, etc.

For the prompt “Write a 200-word LinkedIn post on the topic ‘importance of AI in workplace’. Use short sentences and whitespace for better readability.”, the three components are:

1. Algorithm: Convert the title into a LinkedIn post.

2. Problem statement: The title “Importance of AI in the workplace”.

3. Output description: 200 words. Short sentences. Whitespace.

If you put 1 and 3 in the custom GPT, all you have to type in is the title and you will get a nice, crisp 200-word LinkedIn post.

That’s it. That’s the whole process. But do keep the aforementioned concepts and principles in mind while creating a custom GPT for your writing needs.

Wrapping up: Broaden your scale of thought

It is better to have three GPTs that work well in a sequence rather than one GPT that pisses you off. The aim is to simplify your overall workflows and processes. The agile philosophy suggests that the easiest way to do a complex task is by breaking it into tiny chunks.

Don’t be afraid to do a bit of prompt engineering where you alter the output description or the algorithm slightly to improve the quality of your output. The GPT-creation interface allows for simultaneous testing, take advantage of it.

—

Next Chapter: Making Money With Gen AI: Can Anyone Do It

Previous Chapter: The Ideal LLM for Content Writing

Index (with Prologue): Content Writing With AI