[Journey] 1 million UVs/month in 12 months using AI generated content. Let's do it!

Status
Not open for further replies.
It's a few years old, with DR around 40, but I'd say half of the DR is bullshit backlinks. Some are legit and there was a site in the same niche there before.

around 13k articles. I'm drip-feeding 200/day (6000/month) and indexing using the Google API.

a little bit over $1 cpm now, so not that great I guess? But we're working on scaling like hell now.

In the future will work on a script that deletes the articles that are not performing and also want to add automated translations to articles that work well.
 
I can officially say I consider this journey 100% successful now. This is my biggest PAA site:

View attachment 213452

The funny thing is that the last update everyone is crying about didn't seem to affect PAA sites.

Our app is multithreaded now with full proxy support and I'm in the process of generating this stuff on around 150 domains with my partner this month. We invested in a bunch of strong/aged/high DR domains.

I will still update this thread with new findings and ideas whenever I get the time. Cheers!
Smooth and smiling faces everywhere, but ruin in their eyes.
 
It's a few years old, with DR around 40, but I'd say half of the DR is bullshit backlinks. Some are legit and there was a site in the same niche there before.

around 13k articles. I'm drip-feeding 200/day (6000/month) and indexing using the Google API.

a little bit over $1 cpm now, so not that great I guess? But we're working on scaling like hell now.

In the future will work on a script that deletes the articles that are not performing and also want to add automated translations to articles that work well.
Only 13k articles? monthly earnings?
 
We are using a custom solution right now, but it's very easy to detect repetitions:

Python:
# check if a string has 2 or more words repeated in a row

import re

strings = [
    "Moissanite is naturally occurring silicon carbide and its various crystalline polymorphs. It has the chemical formula SiC and is a rare mineral, discovered by the French chemist Henri Moissan in 1893. Silicon carbide is useful for commercial and industrial applications due to its hardness, optical properties and thermal conductivity.",
    "Moissanite is naturally occurring silicon carbide and its various crystalline polymorphs. It has the chemical formula SiC and is a rare mineral, discovered by the French chemist chemist Henri Moissan in 1893. Silicon carbide is useful for commercial and industrial applications due to its hardness, optical properties and thermal conductivity.",
    "yes no yes no yes no yes no",
    "yes no yes yes no no yes no",
    ]

def check_strings(strings):
    for string in strings:
        if re.search(r'\b(\w+)\s+\1\b', string):
            print("="*50)
            print("Repetitions:", string)
        else:
            print("="*50)
            print("No repetitions:", string)

def main():
    check_strings(strings)

if __name__ == '__main__':
    main()

Output:

Code:
==================================================
No repetitions: Moissanite is naturally occurring silicon carbide and its various crystalline polymorphs. It has the chemical formula SiC and is a rare mineral, discovered by the French chemist Henri Moissan in 1893. Silicon carbide is useful for commercial and industrial applications due to its hardness, optical properties and thermal conductivity.
==================================================
Repetitions: Moissanite is naturally occurring silicon carbide and its various crystalline polymorphs. It has the chemical formula SiC and is a rare mineral, discovered by the French chemist chemist Henri Moissan in 1893. Silicon carbide is useful for commercial and industrial applications due to its hardness, optical
properties and thermal conductivity.
==================================================
No repetitions: yes no yes no yes no yes no
==================================================
Repetitions: yes no yes yes no no yes no

Sorry I'm not replying to a lot of comments guys, but we're trying to capitalize on this thing and working on over 100 nice websites right now.
have any other coding tricks like this to instantly make the text of your articles better?

when you scrape serps, and they all have the keyword in the title ni the top 5-10 articles, how do you chose how to generate a great title from the existing 5-10?
 
@Sartre
Can you share any info on how you managed to scale it to so many sites in terms of getting approved for Adsense?
 
have any other coding tricks like this to instantly make the text of your articles better?

when you scrape serps, and they all have the keyword in the title ni the top 5-10 articles, how do you chose how to generate a great title from the existing 5-10?
Tricks? Hard to say. I would have to go over the whole app and write a top 10 lol. From the top of my head, the library Yake is freaking fantastic: https://pypi.org/project/yake/ and takes zero resources (not ML).

The title keyword is just the keyword I wanted to target + a clickbaity spintax thingy in the end like:
What are the best lightweight city backpacks [Expert advice]
@Sartre
Can you share any info on how you managed to scale it to so many sites in terms of getting approved for Adsense?
Just a combination of:
* a very good looking, polished theme. Looks almost legit. Tried to put an emphasis on EAT.
* good content with no obvious mistakes, weird characters. There's youtube videos in, custom pictures, bold text, quotes, sources, lists, etc.


I'm sorry if I'm not answering every single question but I don't want to doxx my sites. I'm revealing 90% of stuff anyway.
 
My whole operation is being held up by no pages getting indexed. It is pointless scaling up until this is solved. What are people's solutions to this? I see above that @Sartre you are using the 'Google indexing api' I don't know what it is but heard of it - have avoided google tools wherever I can usually. You are using aged DA domains with backlinks already though so your experience will be a lot different anyway I guess. Can anyone using fresh domains comment on how they are indexing links? Mine are about 9 months to 1 year+ old but had no action before just some twitter links, and a couple pinterest/reddit, as I was using for social media only.

I have tried using the youtube indexing method which I setup a few days ago. No success there so far after about 72 hours. The only thing I have used to index links that has worked so far has been submitting to indexing expert's bhw marketplace indexing service. No, not an affiliate or friend that is just the only thing that has worked so far! I think he is only using the youtube indexing method himself from his previous posting giving away the youtube indexing method so not sure what is missing there for me so far.
 
Last edited:
It's a few years old, with DR around 40, but I'd say half of the DR is bullshit backlinks. Some are legit and there was a site in the same niche there before.

around 13k articles. I'm drip-feeding 200/day (6000/month) and indexing using the Google API.

a little bit over $1 cpm now, so not that great I guess? But we're working on scaling like hell now.

In the future will work on a script that deletes the articles that are not performing and also want to add automated translations to articles that work well.


How do you find the articles that are not performing so you can delete them and is there anything positive doing that?

How much time should it pass until you decide to delete an article?
 
Thanks @Sartre for your sharing method, I completely read all comments in the thread after 3 days, there are lots of state-of-the-art information.

I have some concerns about your app:
1. How to create title and subheading of your auto-generated article?
2. Your app used Levenshtein distance for finding similar text between paragraphs. Did you test alternative algorithm like Spacy cosine similarity or fastText WMD algorithm (WMD = Word Mover's Distance)?
3. Did you use playwright-python package instead of Selenium?
 
Last edited:
For the assisted AI content creation I'm using one of the paraphrasing tools that is commercially available. I dont want to advertise anything specifically here. You could just as well use Quillbot like some people do.

For the fully automated creation (muuuuuch faster, a bit more mistakes) I'm running Pegasus on my own PC. I'm using exactly this: https://huggingface.co/tuner007/pegasus_paraphrase
---
To the other guy - there's no ready made tool to do something like this well. At least I haven't seen one. You gotta build it yourself with your own twists. I can give you advice but I'm not going to release my entire sourcecode. It's also not for sale.
That's a smart move man! Good luck dude. Fck millions, make billions already!
 
unrelated but has anyone experimented with manipulating G's rankings by faking CTR in SERPs?
Yes I used to have an agency and one service we even sold was SBO. Search Box Optimization. Basically between microworkers and bots and other methods we'd get the first result in Google Suggestions to be the clients Brand name. DM if more questions on it. Basically it probably helped some or helped if you're google dancing for first vs someone else who has high authrity, but it's not as 80/20 as just more keywords ranking and more traffic etc.
 
My whole operation is being held up by no pages getting indexed. It is pointless scaling up until this is solved. What are people's solutions to this? I see above that @Sartre you are using the 'Google indexing api' I don't know what it is but heard of it - have avoided google tools wherever I can usually. You are using aged DA domains with backlinks already though so your experience will be a lot different anyway I guess. Can anyone using fresh domains comment on how they are indexing links? Mine are about 9 months to 1 year+ old but had no action before just some twitter links, and a couple pinterest/reddit, as I was using for social media only.

I have tried using the youtube indexing method which I setup a few days ago. No success there so far after about 72 hours. The only thing I have used to index links that has worked so far has been submitting to indexing expert's bhw marketplace indexing service. No, not an affiliate or friend that is just the only thing that has worked so far! I think he is only using the youtube indexing method himself from his previous posting giving away the youtube indexing method so not sure what is missing there for me so far.
G indexing API can index 200 pages/day

I mostly use aged domains with decent backlinks, so even without any indexing APIs it's not uncommon to have 10k pages indexed after 2 months.
How do you find the articles that are not performing so you can delete them and is there anything positive doing that?

How much time should it pass until you decide to delete an article?
I think it's less risk to have a website that doesn't have too many articles. Why keep articles that don't get impressions anyway. After 2-3 months it's quite clear looking at GSC.
Thanks @Sartre for your sharing method, I completely read all comments in the thread after 3 days, there are lots of state-of-the-art information.

I have some concerns about your app:
1. How to create title and subheading of your auto-generated article?
2. Your app used Levenshtein distance for finding similar text between paragraphs. Did you test alternative algorithm like Spacy cosine similarity or fastText WMD algorithm (WMD = Word Mover's Distance)?
3. Did you use playwright-python package instead of Selenium?
1. I already replied twice to this. title is the main keywords with positive volume. h2s are related PAAs.
2. We're mostly using Yake for keywords + our own model for relevancy now.
3. I haven't. I will try it for another project. Looks interesting.
Yes I used to have an agency and one service we even sold was SBO. Search Box Optimization. Basically between microworkers and bots and other methods we'd get the first result in Google Suggestions to be the clients Brand name. DM if more questions on it. Basically it probably helped some or helped if you're google dancing for first vs someone else who has high authrity, but it's not as 80/20 as just more keywords ranking and more traffic etc.
That's really cool. We're working on our own app that does it.
 
I got over 100 WP sites on a $20 Linode droplet + cloudflare + wprocket



nah it took me 1 evening to write a better algo than Ahref's KD
You have different servers for the text gen app right? then you upload to wordpress remotely? as this small linode doesn't have the power to do textgen at scale?


hmm what kinda algo? you check if certain sites are ranked and if they are it's over 5 or something, what kind of KD weighing to get a rating?
 
Tricks? Hard to say. I would have to go over the whole app and write a top 10 lol. From the top of my head, the library Yake is freaking fantastic: https://pypi.org/project/yake/ and takes zero resources (not ML).

The title keyword is just the keyword I wanted to target + a clickbaity spintax thingy in the end like:
What are the best lightweight city backpacks [Expert advice]

Just a combination of:
* a very good looking, polished theme. Looks almost legit. Tried to put an emphasis on EAT.
* good content with no obvious mistakes, weird characters. There's youtube videos in, custom pictures, bold text, quotes, sources, lists, etc.


I'm sorry if I'm not answering every single question but I don't want to doxx my sites. I'm revealing 90% of stuff anyway.

Haha! I believe I know you from somewhere. Something you mentioned in the above post just clicked.

Been building PAA sites for a year now. My largest one is 182k pages in an extremely competitive niche. With around 120k indexed.

Been using GPT3 + a custom fine tuned GPTNeoX for content generation. Pegasus for rephrasing.

QA Format is very similar to yours as well.

I also harness the qapage schema in some instances.

Setup is very similar to yours. My tweaking and optimization is towards a different direction, though.

I use a different image generation library that gives me better shadows, more sigma control and crisper fonts.

I don't use the Indexing API (at least haven't yet, but may just use it as a test today on a brand new site).

All, but 5 of my sites, are on brand new Domains with zero back link profiles.

Here's a site with its first post 4 days ago (domain regged same day).

Screen-Shot-2022-06-13-at-8-21-04-PM-1.png


No backlinks, brand new domain, some social signals and web 2.0, again very competitive niche.

Happy to connect and brainstorm if you're interested.
 
Haha! I believe I know you from somewhere. Something you mentioned in the above post just clicked.

Been building PAA sites for a year now. My largest one is 182k pages in an extremely competitive niche. With around 120k indexed.

Been using GPT3 + a custom fine tuned GPTNeoX for content generation. Pegasus for rephrasing.

QA Format is very similar to yours as well.

I also harness the qapage schema in some instances.

Setup is very similar to yours. My tweaking and optimization is towards a different direction, though.

I use a different image generation library that gives me better shadows, more sigma control and crisper fonts.

I don't use the Indexing API (at least haven't yet, but may just use it as a test today on a brand new site).

All, but 5 of my sites, are on brand new Domains with zero back link profiles.

Here's a site with its first post 4 days ago (domain regged same day).

Screen-Shot-2022-06-13-at-8-21-04-PM-1.png


No backlinks, brand new domain, some social signals and web 2.0, again very competitive niche.

Happy to connect and brainstorm if you're interested.
How well are your sites performing despite having 0 links?
 
Status
Not open for further replies.
Back
Top