Rate My New Content AI

hreyhrey

Power Member
Jr. VIP
Joined
Jan 9, 2019
Messages
527
Reaction score
240
Hello everyone,
For the past few weeks I've been working on a specific AI use case and I'm finally beginning to see results.
I'd like to see what others here think of the quality of the content and more importantly how it compares to other AI services on the market. (Jarvis, GPT3, etc)

The content is 100% AI created with my only input being the title of the page (I don't even include a structure) and once in a while deleting certain paragraphs. For this example, I'm using the water/fire damage restoration niche.

Example 1 - https://pastebin.com/BJFTyRj3
Example 2 - https://pastebin.com/3EAyhSr6
 
In my modest point of view, it's a well structured article and I find amazing that all was done based in the title input alone.
Have you checked uniqueness/plagiarism?
Powerful tool right there for spamdexing as many sites as you like.
Congrats op
 
Since this was created without any additional input, this is better than any other AI tool available. I’ve created hundreds of articles with Jarvis but it needs a lot of manual input and editing.
 
This looks great, particularly considering that you don't need to provide structure. However, a little manual editing is required. In first link, "Electrical Insurers" paragraph is posted twice without any changes. Such errors can only be corrected with manual editing or through grammar checking tools. Overall, it looks like a good alternative to existing AI writing tools. You should try developing it into a SAAS project.
 
Man, this Article looks very well written. did you check the grammar and plagiarism
 
In my modest point of view, it's a well structured article and I find amazing that all was done based in the title input alone.
Have you checked uniqueness/plagiarism?
Powerful tool right there for spamdexing as many sites as you like.
Congrats op
Man, this Article looks very well written. did you check the grammar and plagiarism
I’m still tweaking the overall program/process and this was the very first test.

I don’t want to jump to any conclusions on the success just yet since I’m a bit worried about plagiarism in its current state.

Any recommendations for plagiarism checkers? I’ve tried googling random portions of the text and can’t find any explicit plagiarism.

How does google “handle” plagiarism? Is it on the hunt for both exact match plagiarism and content spinning? I mean how many times can people really describe “mold” lol.
 
With no input, this is far better than any of the other tools I've tried. Some of the info is incorrect though, but this sort of fact checking is required when working with conent writers too.
Very impressive.
 
I use rytr and even though it does a decent job it still can't do without manual editing
Any recommendations for plagiarism checkers? I’ve tried googling random portions of the text and can’t find any explicit plagiarism.
You can try out 1text or quetext
 
The major problem with all the AI writers is the facts. Since I don't know anything about home construction or roofing, how am I supposed to know as a reader that the information/tips/suggestions mentioned in the article are valid and effective unless a person specializing in home construction has written it. it is difficult for me as a writer also, since I also don't know anything about this niche. So I cannot even fact check it myself. Some users in other forums have said that I should know all the things in my niche since I should research....but I think, even if I research on the internet for that niche, I will not add something new to the article. It will just be a collection of old articles in a summarized form. I think new ideas/tips/suggestions can only be given by a person who has experience and has worked in that particular area. Now the question is that can AI get that sort of experience by absorbing/training all the information on the internet, even information that is unstructured (video, text, images)....I suppose it is possible since there are algos that can be trained on unstructured data....but the hardware needed will be at a quantum level that is yet to come. Also the question remains is that how can I fact check an article if I myself don't know if the methods/ideas mentioned in the article are true or not. @RoyDanino also pointed out that some info is incorrect but it looks to me that all info is correct. These are the problems that AI needs to resolve otherwise the search engines will fill up with regurgitated content by users like me who are subscribed to one of the GPT3 SAAS services. I am using these services to fill up the content on my site but I don't know if the info is correct or not and I hope big G also does not know! ;)

I would love the idea, if someone comes up with an AI service here to just fact check the articles.
 
Congrats OP.
For AI text generator, the results are impressive. Text is well readable.
I have checked several paragraphs for plagiarism and they were found to be 100% unique. Awesome!
How does google “handle” plagiarism? Is it on the hunt for both exact match plagiarism and content spinning? I mean how many times can people really describe “mold” lol.

As far as I know, Google allows plagiarism, but below 10-20%. Higher percentage can be banned/deindexed.
And one word ("mold" in your example) density can't be higher than 4%.

Please tell me, did you train model or scrape the SERP?
 
the content is Plagiarized
How did you test that?
Congrats OP.
For AI text generator, the results are impressive. Text is well readable.
I have checked several paragraphs for plagiarism and they were found to be 100% unique. Awesome!


As far as I know, Google allows plagiarism, but below 10-20%. Higher percentage can be banned/deindexed.
And one word ("mold" in your example) density can't be higher than 4%.

Please tell me, did you train model or scrape the SERP?
That was the same "plagiarism test" I did. I'm looking into tools now, but so far so good. Well to train a model you need data so why not both?



Plagiarism Update:
So I ran the text through Quetext and the results seem promising. They indicated 13% was "plagiarized", but when you look into the details its very weak evidence and not real plagiarism IMO. I'd love some feedback on this "plagiarism" and whether Google would penalize this content.

1639410086753.png
1639410114091.png
1639410155849.png
 
Last edited:
Example 1 passes the GPT output detector, example 2 doesn't. https://huggingface.co/openai-detector/
I'd say the first example scans better to the human eye too.
Awesome tool, thanks.

It’s an interesting discussion point. Does google care if content is AI created if it
1) fulfills user search intent
2) is not plagiarized
Im thinking this is not on the top of the list for Google to check or care about but am curious of others opinions.
 
How did you test that?

That was the same "plagiarism test" I did. I'm looking into tools now, but so far so good. Well to train a model you need data so why not both?



Plagiarism Update:
So I ran the text through Quetext and the results seem promising. They indicated 13% was "plagiarized", but when you look into the details its very weak evidence and not real plagiarism IMO. I'd love some feedback on this "plagiarism" and whether Google would penalize this content.

View attachment 195314
View attachment 195315
View attachment 195316

You can go the wh route and check each article one by one, or...
You can go the spamdexing route and cloak the traffic to an offer(s)
I would take the bh route and stop worrying about the details, specially if you can generate so much content in such quick time.
Just build site after site according to keyword research and get traffic ASAP.
 
Back
Top