- Nov 23, 2020
- 5,930
- 1,706
I've been analyzing content generation with AI for the last 6 months, made different kinds of research and got some helpful insights.
I'll start with something easy to understand and to do.
The biggest problem of AI content is that it looks exactly like AI-generated. People keep on telling the chatgpt to make a content more human-like, but it still fails. Using this little tip we made our content to be more "human-like" and better for 30-40%.
Step 1. Generate part by part, heading by heading. It's essential because the ChatGPT has its own tokens, and it just can't generate the whole article of high quality with a single prompt. Just can NOT. So do it part by part.
Step 2. After each part generated - don't just go to another one. Read it carefully and if you totally don't like it - give the chat a simple additional prompt, some kind of: "The quality of the content is not what I expected to see. Re-write this part, make it more human like".
You can change this kind of prompt, but the key point is that the rewritten part can be better. It may be more interested to read and understand, and not look like typical AI-content. If you do this for 10 texts - you will see at least 3-4 texts to be better than the original part (they will have less % of AI detection with different tools, and more % of unique with plagiarism checkers)

Have you tried so?
I'll start with something easy to understand and to do.
The biggest problem of AI content is that it looks exactly like AI-generated. People keep on telling the chatgpt to make a content more human-like, but it still fails. Using this little tip we made our content to be more "human-like" and better for 30-40%.
Step 1. Generate part by part, heading by heading. It's essential because the ChatGPT has its own tokens, and it just can't generate the whole article of high quality with a single prompt. Just can NOT. So do it part by part.
Step 2. After each part generated - don't just go to another one. Read it carefully and if you totally don't like it - give the chat a simple additional prompt, some kind of: "The quality of the content is not what I expected to see. Re-write this part, make it more human like".
You can change this kind of prompt, but the key point is that the rewritten part can be better. It may be more interested to read and understand, and not look like typical AI-content. If you do this for 10 texts - you will see at least 3-4 texts to be better than the original part (they will have less % of AI detection with different tools, and more % of unique with plagiarism checkers)

Have you tried so?