My Journey to Automating Reddit Marketing with ChatGPT API and Python. Requesting everyones participation

Paranoid Android

Elite Member
Joined
Jun 20, 2010
Messages
2,864
Reaction score
4,209
Assuming you have a site that you need targeted traffic for, and you've identified the right subreddits to push it to, you're going to need unique human written content in tones matching the questions/posts you're replying to. We as humans having to deal with a lot of things cannot always maintain the mood to make such posts in high volumes, but suprisingly this AI thing is able to do that.

My plan is to make a bot that reads latest posts from a sub that you've picked, read all replies and see if it has your primary keyword occurirng more than a set number of times, and if yes, make a post matching the tone of the others and sneak your link in.

I have 5 aged reddit accounts and a bunch of websites with articles on different niches with adsense. I am going to be using the latest ChatGPT API and the PRAW library for Python to see if this is going to work.

As the development progresses, the code will be updated on this colab Notebook
https://colab.research.google.com/drive/1JJb6JiIxDwcpaJXbTmggugi2lXpR90Do?usp=sharing
Follow me along as I code this up. I request ideas from everyone on how to pick the right posts to respond to, how to decide if the bot should reply to the main post or one or more comments. Let us make something that everyone would benefit from. Non spammy content for the subs, traffic for us, and revenue for the leeches who would be offering this as a service commercially.
 
Last edited:
As a follower from the AI autoblog empire, I have no doubt in OP's ability.
 
Baby Step 1
Initial Prompt:
{"role": "system", "content": "You are an unhelpful sarcastic old man. Keep your responses short and to the point and don't mention anything about various factors. Add the link somewhere in the middle of your response your response : http://somelink.com/"}

Question:
How do you go about nature photography in Bali?

Response:
"Oh, yes, nature photography in Bali is just awful. I'm sure you'll have a blast taking pictures of all the smog and trash that litters the island. And don't even get me started on the overcrowded tourist hotspots. But if you still insist, you can check out somelink.com for some tips and tricks on how to capture the misery in Bali."

Not updated on the notebook yet, but locally I can pull all titles that are questions from a subreddit, display that question to the user so he can decide weather he wants to have that answered, and if yes the q gets passed to openai and the response is again displayed for the users approval, and if approved, it gets posted.

Going Forward towards complete automation:
  • Going to be implementing some tone recognition model to check if it is a real question thats on the title and not a wtf question.
  • For titles that are not questions, the text and comment would be scanned for multiple occurrences of the keyword. The ability to pull a random link from a list of multiple links or maybe even integrate a bunch of urlshortner apis so that the links are always unique.
  • Find questions in comments and reply to those
  • Filter out the 'As an AI model I can't do this shit' sort of replies.
  • Make chatty posts and give useful answers without links for the karma
  • Respond with different tones from different accounts. I went with the unhelpful sarcastic old man cos thats what my dad was. I would appreciate suggestions on other tones I could use. Then again, this is something that the user can play with and modify later.
I shall update the code on the colab notebook in a few hours after testing it. It is still manual and not fully automated. If anyone comes up with more suggestions I could integrate those too.

Might need GPT-4 to avoid ban. ChatGPT text is obvious
not if you prompt it right

As a follower from the AI autoblog empire, I have no doubt in OP's ability.
thank you

There is no fucking way I am making accounts. Too proud and dignified to get into that sort of shit. You're going to have your own accounts.
 
jesus... you're on fire with this chatgpt bot thing... so many ways to monitize. I really regret not taking coding seriously while in high school
 
A smug left leaning neck beard that always needs to be right and goes through other users post history to make a point
what? hows that a tone? did you try it on chatgpt yourself?
 
And the code is updated on the Colab Notebook

Basic instructions in there already.

I've tested it, made one post and this is what you get

Code:
https://www.reddit.com/r/bali/comments/11ir4jp/comment/jb0l26f/

Treat this a v0.1. More updates coming soon. Requesting inputs too.
 
You are putting AI to so many uses. Good luck.
 
Back
Top