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

[Sartre]

First prototype of a live article. This is the high-quality AI. At the moment I can generate maybe 10-20 like this every night only. I'm working on optimizing the process. I will not use this method for spam blogging, as it's way too slow. I think the results are better than cheap writers already:

https://i.imgur.com/pka6WQl.png




Grammarly premium plagiarism returned just some sentence fragments unrelated to running:


(The headings are exact PAAs in this case so they can appear as plagiarism if other sites target them as well)

We're almost there.



The next step is automated linking to studies and sources, formatting text better with <strong> and <I> in some places, and possibly working more on turning photos into illustrations using ML.

PS I'm not publishing this one so that people don't discover my domain.

 

[Mrwhopy]

Impressve! Best of luck!

 

[huhuhu85]

I'm following, all the best!

 

[rampzeus]

Interesting journey.

Do you randomly pick a paragraph or do you use a method for this? Also what do you mean with use paragraph for semantically relevant keywords, does this mean you scrape the paragraph because it's a relevant keyword?

Anyway good luck, will follow this.

 

[Niffo]

The next step is automated linking to studies and sources, formatting text better with <strong> and <I> in some places, and possibly working more on turning photos into illustrations using ML.

By linking to sources do u mean the top 10 ranking sites u scraped. Or do u link to real authority sources and studies? And if so how do u find such studies at scale

Have to say I'm really enjoying this journey. I also hate hiring writers and would rather code in python all day. Ur journey is giving me lots of ideas to try, getting more excited every post already trying some of the libraries to use them in my exciting projects

 

[ANewBeginning]

It's not based on GPT-3 and it's not Quillbot.

is it gpt-2? or some other gpt variations out there?

 

[Sartre]

Interesting journey.

Do you randomly pick a paragraph or do you use a method for this? Also what do you mean with use paragraph for semantically relevant keywords, does this mean you scrape the paragraph because it's a relevant keyword?

Anyway good luck, will follow this.

Look at the structure of my article. I get a main keyword like "running knee pain" and find sub-topics. The for each subtopic I'm looking online for a paragraph that ranks and answers the question.

By linking to sources do u mean the top 10 ranking sites u scraped. Or do u link to real authority sources and studies? And if so how do u find such studies at scale

Have to say I'm really enjoying this journey. I also hate hiring writers and would rather code in python all day. Ur journey is giving me lots of ideas to try, getting more excited every post already trying some of the libraries to use them in my exciting projects

Nah, I'm thinking about looking for actual studies. For health it's easy, because we have PubMed, but for other stuff I don't really know how to pull this off.

is it gpt-2? or some other gpt variations out there?

I don't think even gpt-4 would be able to handle this. I'm not doing text generation, only paraphrasing.


Check these:

mvorganizing.org
dengenchronicles.com
sidmartinbio.org
easierwithpractice.com
janetpanic.com
sluiceartfair.com
idswater.com
colors-newyork.com
holidaymountainmusic.com
restaurantnorman.com
greedhead.net
rehabilitationrobotics.net
everythingwhat.com
askinglot.com

people are banking HARD on AI-generated sites... I hope I'm not too late to the game before Google penalizes this.

The first site got to 5mil visits/month in 6 months

 

[Niffo]

I'm amazed by the quality of content. It's just as good, if not better compared to cheap writers. Also images, categories and external links make allot of sense. AI content is definitely the future

 

[bandititf9]

Look at the structure of my article. I get a main keyword like "running knee pain" and find sub-topics. The for each subtopic I'm looking online for a paragraph that ranks and answers the question.

Nah, I'm thinking about looking for actual studies. For health it's easy, because we have PubMed, but for other stuff I don't really know how to pull this off.

I don't think even gpt-4 would be able to handle this. I'm not doing text generation, only paraphrasing.


Check these:

mvorganizing.org
dengenchronicles.com
sidmartinbio.org
easierwithpractice.com
janetpanic.com
sluiceartfair.com
idswater.com
colors-newyork.com
holidaymountainmusic.com
restaurantnorman.com
greedhead.net
rehabilitationrobotics.net
everythingwhat.com
askinglot.com

people are banking HARD on AI-generated sites... I hope I'm not too late to the game before Google penalizes this.

The first site got to 5mil visits/month in 6 months

How do you know all of them are using AI ? is there specific key points that reveal that?

 

[Sartre]

How do you know all of them are using AI ? is there specific key points that reveal that?

I'm 99% sure. Once you deal with paraphrasers/GANs a lot you get a good feeling for the mistakes they make. Read some of the articles. There are very typical mistakes that AI makes like mixing lbs and kgs in one sentence, etc.

Also, it's not possible to write 5million articles in 4 months

The entire English WikiHow is like 250k articles over 15 years of work of thousands of people.

Edit:
add 2 more to the list above:
ici2016.org
sluiceartfair.com

 

[lastbencher]

Look at the structure of my article. I get a main keyword like "running knee pain" and find sub-topics. The for each subtopic I'm looking online for a paragraph that ranks and answers the question.

Nah, I'm thinking about looking for actual studies. For health it's easy, because we have PubMed, but for other stuff I don't really know how to pull this off.

I don't think even gpt-4 would be able to handle this. I'm not doing text generation, only paraphrasing.


Check these:

mvorganizing.org
dengenchronicles.com
sidmartinbio.org
easierwithpractice.com
janetpanic.com
sluiceartfair.com
idswater.com
colors-newyork.com
holidaymountainmusic.com
restaurantnorman.com
greedhead.net
rehabilitationrobotics.net
everythingwhat.com
askinglot.com

people are banking HARD on AI-generated sites... I hope I'm not too late to the game before Google penalizes this.

The first site got to 5mil visits/month in 6 months

most of them are not spinning or using AI for contents just random scraped paragraph or whole article with same featured images too.. you have achieved far better result. I m doing similar like you, producing avg 10 contents/day , manually checking them and posting. Posting can be automated easily but AI involvement making content generation slow.

 

[Sartre]

most of them are not spinning or using AI for contents just random scraped paragraph or whole article with same featured images too.. you have achieved far better result. I m doing similar like you, producing avg 10 contents/day , manually checking them and posting. Posting can be automated easily but AI involvement making content generation slow.

Still 10/day is much better than the usual 1/day people do. How's your traffic going? Are you doing keyword research?

 

[blogmen]

Nice goal! I love splog journeys. This one is nice because of its simplicity: just paraphrasing content.

Definitely a quality paraphraser.

What are you using to cache posts, just a plugin?

As far as Google banning generated content, do you think it's easy to detect such content? Maybe low quality paraphrasing or spun content. Or they can filter new blogs that post too much too fast.

Will follow this for sure. Looking to build one myself. Good luck!

 

[Machairodont]

Interesting approach!

I was thinking about similar stuff myself and already was playing around with topic/keyphrase extraction for categorizing blog articles into categories.

One of the main obstacles I see is the higher and intermediate hierarchy structure of an excellent human written article.
(just a very simple example)

• intro
• main body
• conclusion

But this varies with the type of article. Here is quite a good overview: https://syedbalkhi.com/how-to-structure-a-perfect-blog-post-with-examples/

Each type of article is structured differently, and of course not every human writer is following these "guidelines"...
If you "just" take paragraphs from multiple articles (maybe even from different types) it will bring you strange results.

You have chosen the "list type" which is, I guess, a good starting point as it consists mainly of lists of facts.
But to have only list type articles on your sites would be kind of boring...

We should have a way to categorize the different spidered articles, to analyze the higher and intermediate hierarchy structure,
and to assemble paragraphs from the same category type and hierarchy level from different articles into a new one,
and this without breaking the context.

Paraphrasing, spinning, putting nice looking pictures and formating everything nicely becomes then the easier part of that effort.

Maybe we should have a chat about that stuff?

Cheers

 

[Skroll]

Very interesting. I'm following

 

[lastbencher]

Still 10/day is much better than the usual 1/day people do. How's your traffic going? Are you doing keyword research?

I started doing this few days back. I m using an expired domain which is idle for long with 3 post only and home page. I m testing domain strength with all type of articles. Published long tails as well as high competition keywords.

 

[Sartre]

Nice goal! I love splog journeys. This one is nice because of its simplicity: just paraphrasing content.

Definitely a quality paraphraser.

What are you using to cache posts, just a plugin?

As far as Google banning generated content, do you think it's easy to detect such content? Maybe low quality paraphrasing or spun content. Or they can filter new blogs that post too much too fast.

Will follow this for sure. Looking to build one myself. Good luck!

Generators like GPT-3 are easy to detect. There are already models for doing so. I'm not sure about the stuff I'm doing.

Interesting approach!

I was thinking about similar stuff myself and already was playing around with topic/keyphrase extraction for categorizing blog articles into categories.

One of the main obstacles I see is the higher and intermediate hierarchy structure of an excellent human written article.
(just a very simple example)

• intro
• main body
• conclusion

But this varies with the type of article. Here is quite a good overview: https://syedbalkhi.com/how-to-structure-a-perfect-blog-post-with-examples/

Each type of article is structured differently, and of course not every human writer is following these "guidelines"...
If you "just" take paragraphs from multiple articles (maybe even from different types) it will bring you strange results.

You have chosen the "list type" which is, I guess, a good starting point as it consists mainly of lists of facts.
But to have only list type articles on your sites would be kind of boring...

We should have a way to categorize the different spidered articles, to analyze the higher and intermediate hierarchy structure,
and to assemble paragraphs from the same category type and hierarchy level from different articles into a new one,
and this without breaking the context.

Paraphrasing, spinning, putting nice looking pictures and formating everything nicely becomes then the easier part of that effort.

Maybe we should have a chat about that stuff?

Cheers

I would argue that paraphrasing is much harder than the other stuff you described. I will add stuff like intros, summaries, other types of articles, silos & interlinking later.

Also about this: "But to have only list type articles on your sites would be kind of boring..." - 95+% visitors of your site will only see 1 article and will come from Google. It doesn't matter if your site is boring. Users are there to click in the ad, not to be entertained.

I started doing this few days back. I m using an expired domain which is idle for long with 3 post only and home page. I m testing domain strength with all type of articles. Published long tails as well as high competition keywords.

Any signals yet?

 

[Sartre]

Nice goal! I love splog journeys. This one is nice because of its simplicity: just paraphrasing content.

Definitely a quality paraphraser.

What are you using to cache posts, just a plugin?

As far as Google banning generated content, do you think it's easy to detect such content? Maybe low quality paraphrasing or spun content. Or they can filter new blogs that post too much too fast.

Will follow this for sure. Looking to build one myself. Good luck!

oh and about the cache, I found a gem - https://wordpress.org/plugins/wp-cloudflare-page-cache/ - it enables full caching for WP without paying for higher tiers of CF and with no problems with dynamic content. My pages load like they are static HTML.

I use it with wp-autooptimize and wp-cache enabler. I also use Optimus HQ Pro for image optimization.

 

[bandititf9]

oh and about the cache, I found a gem - https://wordpress.org/plugins/wp-cloudflare-page-cache/ - it enables full caching for WP without paying for higher tiers of CF and with no problems with dynamic content. My pages load like they are static HTML.

I use it with wp-autooptimize and wp-cache enabler. I also use Optimus HQ Pro for image optimization.

Have you tried WP Rocket for optimization ?

 

[Sartre]

Have you tried WP Rocket for optimization ?

I haven't yet. I will try it though.