I am trying to train the GPT to know about my business, the target audience and so on to always remember and keep the information in mind when writing SEO optimized content for my blogs.
However the instruction field and the conversation starters are confusing me. I thought that I could just make a pdf explaining how I want the blogs to be done specifically, but now there is a seedling that I might have to do that in the instruction field. Does it matter or should I instruct it to write as explained in document X in there?
And then maybe an optional quality of life question:
How would I go about making that a flow where a conversation starter would be Write a Blog Post and then we go through an iterative list for example Topic. Then he asks me for keywords or however i set it up until the prompt is fully crafter and then executes. How would one go about that?
I hope I could convey my thoughts well enough and thanks for the answers in advance
Hello, I can give you my thoughts about your question, because I’m using custom GTPs to write blog posts for diferent topics.
In the PDF section I just gave the GTP the layout of the blog post. You can even add more types of blog layouts depending on the topic.
In the instructions section I explained what I wanted the GTP to do step by step, when I want it to create a blog post. and I gave strict instructions to do that. Here is an example
" when I ask you to create a blog post you will do as fallows:
first you will research online for trending topics
you will then give me 10 blog post titles
I will chose 1 blog title and you will write the blog post using the layout I provided in the “xwz” document
the blog post must be 1000 words long
the blog post must include SEO keywords in the title and first section of the blog posts
after you write the blog post you will give me a list of seo keywords
you will provide a 160 letter snippet including the seo keywords"
Keep in mind that the instructions section has limited word count.
Hopefully this helps. This method works for me and I have 5 diferent writters for diferent topics for my blog. You can give the writters diferent tones, expertise, etc.
There are several ways to go about this. You can write several different prompts as “modes” and put them into a PDF, then write instructions to tell the Custom GPT to use the PDF instructions for each mode and leverage these specific instructions for each phase of your project. That way, you can change “modes” within the same GPT and still have the info context between each step. I’ll call this Option 1.
Another way you can do this (Option 2) is to write a custom GPT per task/phase of your work. Then start a “master” GPT where you “call in” your first Custom GPT as you start your task and then call in the other GPT’s as you progress thru the project. I believe ultimately this is the best way to go because it keeps your workflow simpler and it utilizes the latest features of ChatGPT. Down the road OpenAI may even allow us to automate these steps for quicker outputs.
Here is an example of an experiment I created to test Option 1 using the PDF instruction set:
I’m intending to convert this over to Option 2 eventually to see how well that works.
Thank you! Option 2 probably is the bestas you could fine tune each step through the personas and more in the various steps. As I am a beginner I guess I will try to do all in one