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

[castoro]

I'm avoiding instant indexing from now on.

so you do still connect GSC to the site but do not force the indexing neither with GSC or other plugins, got it! thanks

 

[Sartre]

so you do still connect GSC to the site but do not force the indexing neither with GSC or other plugins, got it! thanks

Yeah I connect GSC I'm just not asking to index.

 

[castoro]

Yeah I connect GSC I'm just not asking to index.

thanks for clarifying btw thanks to you I'm learning python!

 

[hwp994108018]

GPT-3 is not that great to be honest. There are better models out there already in the public. Unless you have a ton of money to fine-tune it.

Also I don't care about Bing.

When I built my first website, I got more Bing traffic than Google traffic, although I'm now also focused on Google. But it's a shame to give up bing, bing has a dedicated WordPress plugin to help webmasters submit links, which is convenient, you should probably give it a try. Plugin: indexnow

 

[hwp994108018]

the whole point is guiding the app into the best low competition PAAs

low competition PAAs is everything

 

[Sartre]

low competition PAAs is everything

right now there's a 100 other things you have to get right to avoid penalties. You can't just blast articles.

 

[FrozenSpade]

right now there's a 100 other things you have to get right to avoid penalties. You can't just blast articles.

Can you share some of them? My legit, PAA-style site got penalized and I am still trying to find out why.

BTW is your tool using GPT3? Or how does content generation work?

 

[BlackHatNoob01]

This is an example of the high-quality paraphrasing. Each sentence takes about 20 seconds on my computer. The fast paraphrasing is about 20x faster, but makes less sense. AI is trying to pick a sentence that is the least similar to the original out of the candidates.

View attachment 187973

Also wanted to show you an example image I'm generating.

View attachment 187975

Damn that looks clean af.

 

[H3llas]

This is an example of the high-quality paraphrasing. Each sentence takes about 20 seconds on my computer. The fast paraphrasing is about 20x faster, but makes less sense. AI is trying to pick a sentence that is the least similar to the original out of the candidates.
...
Also wanted to show you an example image I'm generating.
...

You didn't paraphrased text in the example image? Do you think Google detect texts on images and use it as some sort of relevancy signal?

 

[Sartre]

You didn't paraphrased text in the example image? Do you think Google detect texts on images and use it as some sort of relevancy signal?

Google looks at the big picture. Content, number of articles, content/link velocity and quality. People are looking into details too much. Like duplicate content or Google detecting AI. There is NO WAY that Google is even trying to detect AI. I see a copy-paste site with 1mil indexed articles getting 500k monthly visits after a year. Still there after the last update.

 

[Tuhin Mahmud]

Have you fine-tuned the paraphrasing model?

 

[Sartre]

Have you fine-tuned the paraphrasing model?

I'm not paraphrasing anymore. Pure AI generation right now.

 

[Tuhin Mahmud]

I'm not paraphrasing anymore. Pure AI generation right now.

Thanks for replying, may I ask, any open-source model you're using?

 

[FrozenSpade]

I'm not paraphrasing anymore. Pure AI generation right now.

Does it use gpt3?

 

[H3llas]

Spent last 3 evenings reading the topic from begining.

I have been lurker here for at least 15 years, but your journey and style of writing, talking and thinking is so great that I had to leave a comment.

I am wondering, do you like chess and did you ever played chess in some point of your life?

Thank you.

 

[Sartre]

Does it use gpt3?

I'm using a new model now that I trained after the update on Graphcore IPUs based on this research: https://arxiv.org/abs/2205.14217

 

[Sartre]

It's time.

I'm starting my second journey

https://www.blackhatworld.com/seo/sartres-second-journey-taking-an-ai-site-from-1-000-mo-to-30k-mo-in-2023-lets-go.1454053/​

 

[Sartre]

I've been asked about my keyword research process, and I will elaborate here:

I know this might sound unbelievable, but my keyword research tool is anywhere from 50 to 100 times more efficient than Semrush. In fact, my app is not only in a different league, but we're playing different games.

Sorry for sounding cocky, but I've been coding my life away for a year to create this

Let me prove it:

How Semrush operates:

1. They got a database of, let's say, 2B English keywords and their SERPs.
2. They calculate the DA and PA for each page on the Internet that they know of based on their vast backlinks and SEO juice database.
3. You search for "Tennis". They query their keyword database, sort the keywords by the pages with the lowest PA and DA, and calculate the KD this way.
4. MAYBE they do some additional relevancy checks on demand. I haven't used them in a while cause their and Ahrefs KD never worked for me.

---

I don't think KGR is a good indicator of difficulty. Think about a keyword like "female divorce attorney north Hollywood". Low volume, low allintitle, huge cpc, huge competition.

How my app works:

1. I've got a database of 1B English keywords.
2. I will query the word "tennis". Let's say I will get 10k results.
3. Now my 48 core server running 300 threads/proxies will scrape up to 200 People Also Ask, autosuggest, and related keywords. Per keyword. This takes a week. Now, let's say we have 200k unique keywords with the word "tennis" in either keyword or the snippet. Including keywords like "What racquet does Roger Federer use?"
4. Now, I will run an analysis using a SERPs API service to discard all the keywords that have 0 US volume. We're down to 20k keywords.
5. Now, we analyze each of these keywords using the server mentioned above. My app will go and read every article on page 1 to analyze topical relevancy to that keyword/snippet. Also, taking into account domain authority, but not to the extent Semrush does. Bonus points if we find user-generated content. (Reddit, Quora, forums, etc.) This step takes another week.
6. We end up with 2k keywords which are relatively easy.
7. Rinse and repeat if necessary.

This way, even if I pay cash for outsourcing an article, I have a much better chance I will rank high, so my money is spent more efficiently.

For computational reasons, services like Ahrefs or Semrush can't compete with what I'm doing because they have tens of thousands of customers (or hundreds of thousands?) and I have dozens.

 

[Beklink]

I've been asked about my keyword research process, and I will elaborate here:

I know this might sound unbelievable, but my keyword research tool is anywhere from 50 to 100 times more efficient than Semrush. In fact, my app is not only in a different league, but we're playing different games.

Sorry for sounding cocky, but I've been coding my life away for a year to create this

Let me prove it:

How Semrush operates:

1. They got a database of, let's say, 2B English keywords and their SERPs.
2. They calculate the DA and PA for each page on the Internet that they know of based on their vast backlinks and SEO juice database.
3. You search for "Tennis". They query their keyword database, sort the keywords by the pages with the lowest PA and DA, and calculate the KD this way.
4. MAYBE they do some additional relevancy checks on demand. I haven't used them in a while cause their and Ahrefs KD never worked for me.

---

I don't think KGR is a good indicator of difficulty. Think about a keyword like "female divorce attorney north Hollywood". Low volume, low allintitle, huge cpc, huge competition.

How my app works:

1. I've got a database of 1B English keywords.
2. I will query the word "tennis". Let's say I will get 10k results.
3. Now my 48 core server running 300 threads/proxies will scrape up to 200 People Also Ask, autosuggest, and related keywords. Per keyword. This takes a week. Now, let's say we have 200k unique keywords with the word "tennis" in either keyword or the snippet. Including keywords like "What racquet does Roger Federer use?"
4. Now, I will run an analysis using a SERPs API service to discard all the keywords that have 0 US volume. We're down to 20k keywords.
5. Now, we analyze each of these keywords using the server mentioned above. My app will go and read every article on page 1 to analyze topical relevancy to that keyword/snippet. Also, taking into account domain authority, but not to the extent Semrush does. Bonus points if we find user-generated content. (Reddit, Quora, forums, etc.) This step takes another week.
6. We end up with 2k keywords which are relatively easy.
7. Rinse and repeat if necessary.

This way, even if I pay cash for outsourcing an article, I have a much better chance I will rank high, so my money is spent more efficiently.

For computational reasons, services like Ahrefs or Semrush can't compete with what I'm doing because they have tens of thousands of customers (or hundreds of thousands?) and I have dozens.

With Ahrefs raising their prices and people looking for alternatives, its time for Sartre’s SEO tool to go public

 

[Sartre]

With Ahrefs raising their prices and people looking for alternatives, its time for Sartre’s SEO tool to go public

that's very nice of you to say. It's a bit difficult to package this for people. It's a command line tool and the python code would not be obfuscated in any way. I will think about this.