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

Hey folks,

I have a background in computer science. I already own several profitable content websites (but nothing crazy), and I'm tired of creating/outsourcing content.

I've created a simple app in Python that goes through the top results on Google for a given keyword, takes a paragraph from each for semantically relevant keywords, constructs a new article out of it, and paraphrases it using an AI tool. It also generates related images, adds nice formatting, and schema (I'm using FAQ schema a lot for PAA keywords).

I'm using WordPress on Linode with Centminmod. Posting using the REST API.

In the nearest future I'm launching 3 sites:

1. My passion hobby project - I will generate articles using AI, but edit them manually - So far this one is up with 3 articles, started yesterday.
2. A big site where I will drip tens of thousands of posts without editing and try to monetize with display ads.
3. Another fully-automated site that will target local keywords for lead generation.

Attaching a sneak peek of my app. I will show you an example article in my next update.

Wish me luck!

Good luck with your journey!

There's another guy here who went down the same path and did not have any success. You might want to check him up. Not linking so you can do your own browsing, there's plenty to learn in your situation with AI.

You might want to check what he did and figure out what he did wrong, maybe that will help. The concept is the same though.

Care to share the script!? Haha. I'd love to try that. I never thought of creating something like this using Python. Great idea!

Alexion said:

Good luck with your journey!

There's another guy here who went down the same path and did not have any success. You might want to check him up. Not linking so you can do your own browsing, there's plenty to learn in your situation with AI.

You might want to check what he did and figure out what he did wrong, maybe that will help. The concept is the same though.

Click to expand...

Thanks! Which one was that? I've seen another thread where someone was successful: https://www.blackhatworld.com/seo/scaling-autogenerated-content-to-100-000-page-views-a-month.1313311/page-14
My bet is the websites that fail are just low quality, lots of gibberish, bad design, etc.

MauriceHolem said:

Care to share the script!? Haha. I'd love to try that. I never thought of creating something like this using Python. Great idea!

Click to expand...

I won't share the script for the simple reason that I'm using an AI app that doesn't have an official API and I'm breaking the terms of service and want to fly under the radar(this app isn't very well-known). It's not based on GPT-3 and it's not Quillbot.

I can share parts of the code, for example how I create the images. And I can answer any questions and give you tips.

good luck with your journey, following

Which AI tool are you using to paraphrase content?

interesting, following

Can you share any tools name where I can import a huge list of kws and it will generate content automatically?

WIZDAHL said:

Which AI tool are you using to paraphrase content?

Click to expand...

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.

good luck following

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.

Your paraphrasing results are quite impressive, and you even have your own algorithm to select the best candidate. Can you give me some hints on the selection criteria? One of my guesses is counting and comparing the new words on the candidates, is it reasonable?

In fact, are you using the "P" tool on the online notebook or on the python ide? This is because it is very difficult to install the P tool on the ide successfully.. many error message occurs

DaliW said:

Your paraphrasing results are quite impressive, and you even have your own algorithm to select the best candidate. Can you give me some hints on the selection criteria? One of my guesses is counting and comparing the new words on the candidates, is it reasonable?

In fact, are you using the "P" tool on the online notebook or on the python ide? This is because it is very difficult to install the P tool on the ide successfully.. many error message occurs

Click to expand...

I'm just running plain old Python locally using virtualenv, Python 3.6.8 actually, because some of the dependencies (Nider) don't run on newer versions.

At first, I was using Spacy to calculate sentence similarity using a ML model, but it took MUCH more time and yielded just a bit better results than plain old counting of repeated words.

I'm also thinking about adding a function to turning photos from Unsplash into illustrations and make some stuff kinda like WikiHow.

Will it be faster if you run on Colab instead of your own PC or maybe hire some servers.

brrupsz said:

I'm just running plain old Python locally using virtualenv, Python 3.6.8 actually, because some of the dependencies (Nider) don't run on newer versions.

At first, I was using Spacy to calculate sentence similarity using a ML model, but it took MUCH more time and yielded just a bit better results than plain old counting of repeated words.

I'm also thinking about adding a function to turning photos from Unsplash into illustrations and make some stuff kinda like WikiHow.

Click to expand...

Thanks for your sharing. I will try to install the tool on the older python.

You're right, using package will give the better results, but it takes longer than using the simple scripts.

I think it is possible to use adobe api to cartoonize the photo automatically. It seems we can also use the batch tool on the software, so we don't need to write and test the python scripts.

vin42 said:

Will it be faster if you run on Colab instead of your own PC or maybe hire some servers.

Click to expand...

Yes, but I wish to incorporate the codes in my workflow. I don't want to put everything on Colab.

vin42 said:

Will it be faster if you run on Colab instead of your own PC or maybe hire some servers.

Click to expand...

I really want to do stuff locally for now as much as possible. I think I don't really need to generate more than few hundred articles a day if they are really good. I'm aiming for going for quality over quantity. I think that's less chance of getting manual penalties. Will see. If it's too slow I will invest in cloud GPU.

DaliW said:

Thanks for your sharing. I will try to install the tool on the older python.

You're right, using package will give the better results, but it takes longer than using the simple scripts.

I think it is possible to use adobe api to cartoonize the photo automatically. It seems we can also use the batch tool on the software, so we don't need to write and test the python scripts.

Click to expand...

I'm almost done fine-tuning a photo -> illustration GAN using some Chinese whitepaper and their repo. TensorFlow + Keras. I will show the results soon https://github.com/mnicnc404/CartoonGan-tensorflow

brrupsz said:

I really want to do stuff locally for now as much as possible. I think I don't really need to generate more than few hundred articles a day if they are really good. I'm aiming for going for quality over quantity. I think that's less chance of getting manual penalties. Will see. If it's too slow I will invest in cloud GPU.


I'm almost done fine-tuning a photo -> illustration GAN using some Chinese whitepaper and their repo. TensorFlow + Keras. I will show the results tomorrow https://github.com/mnicnc404/CartoonGan-tensorflow

Click to expand...

That's great. I am looking forward to your results.

How many articles have you done so far, as in published?

This is going to be epic.
Will follow to learn something new.
Best of luck op!