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

[GVald]

Well, yes, but it's a great starting point for many.

Personally, I use Selenium without any proxies. One request can fetch even 300 questions.

I do this aswell. Just helping out

 

[neOtron]

Well, yes, but it's a great starting point for many.

Personally, I use Selenium without any proxies. One request can fetch even 300 questions.

one query 300 questions wow!
how???
i was under impression that in one page there are onlt 4 questions. once we click the last question then 2 more question will show up. and i thought that clicking that last question will also count as a new query.
please enlighten me how to get 300 questions in one go

 

[Alexion]

one query 300 questions wow!
how???
i was under impression that in one page there are onlt 4 questions. once we click the last question then 2 more question will show up. and i thought that clicking that last question will also count as a new query.
please enlighten me how to get 300 questions in one go

You get 4 questions initially.
From the last to the first, click on each, once.
Another 12 will appear. Repeat the above process. Scroll into view each time.

 

[neOtron]

You get 4 questions initially.
From the last to the first, click on each, once.
Another 12 will appear. Repeat the above process. Scroll into view each time.

Ok understood. I knew this but I never thought about doing it as I presumed that it will fill in all my requests bucket as When we click any question it calls a async request to Google. I guess I was wrong. Thanks again

 

[neOtron]

Not really possible to make it perfectly yet imo. Even grammarly makes some bullshit mistakes often.

i am currently experimenting with a few libraries . there are giving descent result. perhaps you might already have tested them but mentioning them here as this thread is helping me alot so i should also share my findings as it might help others. Here it goes.

When i collect texts from different source, or paraphrase then most of the time i loose all the punctuations and texts become very unsegmented like missing fullstops(periods), commas, question mark, and get a lot of wrong spellings etc. to solve this issue i am first using
the library
https://github.com/notAI-tech/fastPunct/
and
https://github.com/ottokart/punctuator2
this above two libraries in combination will add all the missing fullstops, commas, question mark . and create a better looking segmented article.

on top of this i am testing with this grammer correction library : https://github.com/jxmorris12/language_tool_python it basically similar to grammerly but less accurate ofcourse. it also automatically corrects alot of grammer mistakes.

my code is nowhere near perfect and i m still trying to find better libraries . please let me know if you guys have recommendation about some more libraries

 

[Shaybib]

i am currently experimenting with a few libraries . there are giving descent result. perhaps you might already have tested them but mentioning them here as this thread is helping me alot so i should also share my findings as it might help others. Here it goes.

When i collect texts from different source, or paraphrase then most of the time i loose all the punctuations and texts become very unsegmented like missing fullstops(periods), commas, question mark, and get a lot of wrong spellings etc. to solve this issue i am first using
the library
https://github.com/notAI-tech/fastPunct/
and
https://github.com/ottokart/punctuator2
this above two libraries in combination will add all the missing fullstops, commas, question mark . and create a better looking segmented article.

on top of this i am testing with this grammer correction library : https://github.com/jxmorris12/language_tool_python it basically similar to grammerly but less accurate ofcourse. it also automatically corrects alot of grammer mistakes.

my code is nowhere near perfect and i m still trying to find better libraries . please let me know if you guys have recommendation about some more libraries

Can you explain more you use both libs combined? I didn't really get what the second library does...
For spelling you can also use gingerit python library.

 

[neOtron]

Can you explain more you use both libs combined? I didn't really get what the second library does...
For spelling you can also use gingerit python library.

Isn't gingerit just a wrapper for a third party service? I saw its code and it needed api keys so I dint play with it. Am i wrong ? Please enlighten me.

I m combining those two punctuation libraries bcos first of all I have unsegmented texts (why I have these unsegmented texts it's bcos I am getting articles from some source which I can't mention) which has no full stops and commmas. If i feed 2000 words of texts in one go to any of the libraries my 32 gb Ram server starts crying and mostly freeze up. So I break the text randomly in chunks of 200 or so words and then run them in both libraries and combine the results . As both libraries work in different concepts and different datasets so their results are different.

Edit: I just use python as standalone webservices on my servers as I m not really good at python. All my front end ,backend and most of the data processing is done at php and my php scripts calls the python services . As I don't want to spend alot of time in learning too much python so I intend to use it as standalone service APIs untill I achieve some result then after that I will think about optimization by converting all processing and db queries in python

 

[Shaybib]

Isn't gingerit just a wrapper for a third party service? I saw its code and it needed api keys so I dint play with it. Am i wrong ? Please enlighten me.

You are right.

 

[prvncqr]

@Sartre How much do you make daily?

 

[Rondi8754]

Good idea. I'm following...

 

[Sartre]

If you were ready to spend $200 for API, you could have bought 100% private proxies from a legit site and scraped millions of sites in a month easily by just doing a small correction in your script.

It's more efficient to pay for the API and have the results within a few minutes for 20k keywords that I need at that point

i am currently experimenting with a few libraries . there are giving descent result. perhaps you might already have tested them but mentioning them here as this thread is helping me alot so i should also share my findings as it might help others. Here it goes.

When i collect texts from different source, or paraphrase then most of the time i loose all the punctuations and texts become very unsegmented like missing fullstops(periods), commas, question mark, and get a lot of wrong spellings etc. to solve this issue i am first using
the library
https://github.com/notAI-tech/fastPunct/
and
https://github.com/ottokart/punctuator2
this above two libraries in combination will add all the missing fullstops, commas, question mark . and create a better looking segmented article.

on top of this i am testing with this grammer correction library : https://github.com/jxmorris12/language_tool_python it basically similar to grammerly but less accurate ofcourse. it also automatically corrects alot of grammer mistakes.

my code is nowhere near perfect and i m still trying to find better libraries . please let me know if you guys have recommendation about some more libraries

this is really cool and impressive.

Right now we are working on a theme that is very flexible/randomized with different color pallettes and looks like it's a very legit site like Very Well Health etc. trying to target all EAT factors from the Google manual review handbook https://static.googleusercontent.com/media/guidelines.raterhub.com/en//searchqualityevaluatorguidelines.pdf

Right now scaling and improving the system is my goal, not making pennies. My goal is to get these sites into Mediavine/Adthrive, with slight editing. My white hat sites are already making a full-time income so I'm not pressed for making money. I wouldn't monetize a site that is getting less than 50k US impressions/month. I tried and got some sites into Ezoic just as a proof of contept but it's small money and a shitty network so we removed these ads.

 

[prvncqr]

It's more efficient to pay for the API and have the results within a few minutes for 20k keywords that I need at that point



this is really cool and impressive.

Right now we are working on a theme that is very flexible/randomized with different color pallettes and looks like it's a very legit site like Very Well Health etc. trying to target all EAT factors from the Google manual review handbook https://static.googleusercontent.com/media/guidelines.raterhub.com/en//searchqualityevaluatorguidelines.pdf

Right now scaling and improving the system is my goal, not making pennies. My goal is to get these sites into Mediavine/Adthrive, with slight editing. My white hat sites are already making a full-time income so I'm not pressed for making money. I wouldn't monetize a site that is getting less than 50k US impressions/month. I tried and got some sites into Ezoic just as a proof of contept but it's small money and a shitty network so we removed these ads.

Ok. How many posts are indexed currently? And, what tool do you use to check the keyword difficulty level?

 

[dpquy2020]

Scraping google isn' that hard from my experience. Can run 6 concurrent scripts scraping google on one good ip. For insance a 4/5G connection from your phone. Scraped arround 60K serps a day, not a single captcha. Can easily scale with more 4G sims and laptops. Also if u live near public wifi, u can abuse those. Those ip's have a great reputation because of all the real users using them, bet u can easily run 20+ concurrent scrapers on there.

It is so great. do you use selenium or request?

 

[siscooo]

How long(s-ms) it's taking paraphrasing per sentence with your current stack(if you are using cuda)

 

[desperado596]

@Sartre Are you using Wordpress ? what is gonna happen when too much post published ? 500k or 1 million.
Can wordpress still function ? i would expect crush not loading websites and db crush. plesk 404 errors etc.
how many post did you published until now ?

 

[Sartre]

Ok. How many posts are indexed currently? And, what tool do you use to check the keyword difficulty level?

tens of thousands across domains

How long(s-ms) it's taking paraphrasing per sentence with your current stack(if you are using cuda)

I'm publishing around 100-200 articles/day right now, but they are high quality. If you skim they look like they are written by a human.

@Sartre Are you using Wordpress ? what is gonna happen when too much post published ? 500k or 1 million.
Can wordpress still function ? i would expect crush not loading websites and db crush. plesk 404 errors etc.
how many post did you published until now ?

Posting 1million posts is too risky imo. I want to my sites to stay up long-term.

Yes WP will function. I don't use any GUI managers like Plesk, but I don't see why they would be a problem. I've had a site with around 80k posts about nutrition, load testing 250 concurrent visitors for 1 hour, no problems.

I host sites on Vultr high frequency, Centminmod, WP-rocket(turns out it's really good now), wp super page cache, cloudflare. This is my current stack, I'm getting 100 mobile scores with Mediavine on.

 

[Sartre]

What I've been working on:

• better images
• internal linking
• external linking
• automatic seo-friendly translations to other languages with their own translated urls using the Deepl API. Inspired by - https://midogguide.com/
• making articles look like they are written by a human. lists, boldings, italics, youtube embeds

 

[Shaybib]

Any pointers on how to make automatic interlinks? I struggle with it very hard.

 

[Sartre]

Any pointers on how to make automatic interlinks? I struggle with it very hard.

site:yourdomain.com %new_keyword%

Google tells you where you should link to in order of relevance (on established sites)

 

[Archemike]

thanks! I'm only using Newspaper for competition analysis right now. Always improving my app.

yup Python all the way.

some easy keywords take days to rank

~1000 words per article. They pretty much pass as low quality human articles.

Parrot is meh, for example because it loses all the capitalizations and some special characters. Pegasus is also meh but it can be fine tuned quite well, but that requires some knowledge and money.

it's possible to reverse engineer their private API that is used for their chrome extension but that requieres major skills. If someone can do this, message me.

Nice catch!! It's an absurdist joke, I thought that if Camus would make an account on BHW he would call himself Sartre with whom he was always arguing.

Also it sucks a bit when you're doing it at a scale cause there's some random server errors, screw ups at their side once in a while. A pain in the ass to be honest doing this with Selenium.

---

At this point my app is really robust and creating auto sites. I only need to set up the domain, REST API, choose topic, categories and sub-categories.

View attachment 197960

one of the auto sites. I'm 95% certain at this point that the goal of this journey will be met in max. 6 months.

you can use a MITM to map their private api. is quillbot only one with this? does anyone now if rytr or wordtune has a private api too? could find the routes easily, only problem if you consume too much they'll see the requests in their servers and might wonder why ip addresses that aren't internal are consuming so much of these endpoints