Very impressive. I am interested to see it. Enough people have with me in my team.Hello everyone.
To reference my e-Commerce site I created an AI that can help me and save time.
I know that many of you have the same worries as me, that is to say that they spend a lot of time trying to make quality content to please google.
To help me, I created an AI that allows me to :
- retrieve the search ideas associated with my keywords, sort them, associate them with a page, etc.
- generate an article structure, so how many H2, H3 and P etc.
- generate content following the previous structure
- generate synonyms to avoid plagiarism.
The results are not perfect and have to be edited by hand, but I save a lot of time, and I don't lack ideas anymore!
What we are most interested in is the performance:
In my GSC, out of 1000 keywords referenced, 696 are in the first 5 positions.
By the way, if some people are interested, let me know, if there are enough people, I will put it online.
Interesting concept. Would like to know moreThank you all for your interest.
So I'm making a graphical interface a bit nicer than the one I have now. I think I could be done in 2-3 days.
As soon as I finish the first pages I'll make you a screenshot so you can see how it looks like.
As far as how it works.
No I don't use GPT-3. First of all there are several solutions that exist for that and some with very low prices, so it would have been a waste of time for me to do the same thing again.
Secondly, the goal of GPT-3 is to make text looking like written by a human. My goal is to make content that is well referenced by google, I don't care if it looks like human text or not.
Google uses a Latent Dirichlet Allocation algorithm. This allows to sort documents by subject. Basically you need a large number of documents, have them analyzed by this algorithm and it will recognize the ideas and themes, and deduce a kind of profile by subject.
So if we do the reverse principle, we search a subject on google, we analyze the profiles of the best referenced sites to deduce a typical profile, and the AI proposes a result that takes into account this profile.
The beauty of this algorithm is that the profile does not depend on words, but on ideas and themes. That's why google says not to try to repeat your target keyword several times, and to write as naturally as possible.
So in our case, if we get the profiles of ideas that Google liked, and we change some words by synonyms, we modify some sentences to make it more readable for the client, we keep the idea and we have original content and not plagiarized.