Results from different AI content generation methods

Linkbutler

Regular Member
Joined
Aug 20, 2020
Messages
467
Reaction score
173
Wanted to show you guys two graphs from search console.
I have two sites, they have pretty much the same stats. I've made a custom tool for content generation, so site has it's own chain prompt flow.

The different is crazy, it honestly proves to me that AI content is fine, but the prompting is extremely important

Site 1:
1.png


Site 2:
2.png
 
wow thats impressive. I created AI content as well and it brought impressions, but the click rate is very low, I think 0.1%. and after about 3-4 months they de-indexed most of my pages and the impressions disappeared.

how do you build it up and ensure real users find it useful?
 
I agree that this is a great example of how prompting and content generation workflows can make a huge difference. Both sites have similar publishing methods, yet one is ranking significantly better and earning far more impressions and clicks. It suggests that AI content itself isn't the issue-it's the quality, structure, relevance and usefulness of the output that matter. The tool is only as good as the process behind it.
 
wow thats impressive. I created AI content as well and it brought impressions, but the click rate is very low, I think 0.1%. and after about 3-4 months they de-indexed most of my pages and the impressions disappeared.

how do you build it up and ensure real users find it useful?
I've been testing a lot, think I've spend at least 500 hours in total :D
I also use multiple LLMs to write it (through their apis ofc, doing it manually would be insane)
 
I’ve also had some sites showing this kind of result before. The problem is, once a Google update hits, some of them drop immediately and never come back

I’d be interested to know how old your sites are.
 
I’ve also had some sites showing this kind of result before. The problem is, once a Google update hits, some of them drop immediately and never come back

I’d be interested to know how old your sites are.
Both sites are 1-2 years old
 
Now there’s your big lesson learned: This really does show that Google isn’t murdering AI-generated content; they’re murdering *lazy* AI-generated content.

Care to offer a quick analysis on how this particular chain prompt works differently (semantic structure, data insertion, or perspective switch)? The forum would love some concrete information like that.
 
I also generate content and publish but i have two different results in two different websites.

This one got traction asap it got published
1781339715566.png

This one didn't. I am still working on it.
1781339797149.png

It totally depends how good i am fulfilling users query.
 
Wanted to show you guys two graphs from search console.
I have two sites, they have pretty much the same stats. I've made a custom tool for content generation, so site has it's own chain prompt flow.

The different is crazy, it honestly proves to me that AI content is fine, but the prompting is extremely important

Site 1:
View attachment 528757


Site 2:
View attachment 528758
Nice man. Which AI LLM did you use to create your content?
 
Wanted to show you guys two graphs from search console.
I have two sites, they have pretty much the same stats. I've made a custom tool for content generation, so site has it's own chain prompt flow.

The different is crazy, it honestly proves to me that AI content is fine, but the prompting is extremely important

Site 1:
View attachment 528757


Site 2:
View attachment 528758


can you share some of your prompts with us??
 
can I ask how long the prompts are, and how long the output content is? Just for my own curiosity, you don't have to share if you don't want :)
All content is 1200-3000 words :)
The prompts are quite long and involve multiple steps (creating outline, writing initial draft, proof reading, fact checking etc) so it takes my generation tool 5-10 minutes to generate 1 article
 
All content is 1200-3000 words :)
The prompts are quite long and involve multiple steps (creating outline, writing initial draft, proof reading, fact checking etc) so it takes my generation tool 5-10 minutes to generate 1 article
thanks :)
 
Wanted to show you guys two graphs from search console.
I have two sites, they have pretty much the same stats. I've made a custom tool for content generation, so site has it's own chain prompt flow.

The different is crazy, it honestly proves to me that AI content is fine, but the prompting is extremely important

Site 1:
View attachment 528757


Site 2:
View attachment 528758
Good job.
I'd be interested to know how traffic conversion into $$$ is going, and what monetization model you're using?
 
Now there’s your big lesson learned: This really does show that Google isn’t murdering AI-generated content; they’re murdering *lazy* AI-generated content.

Care to offer a quick analysis on how this particular chain prompt works differently (semantic structure, data insertion, or perspective switch)? The forum would love some concrete information like that.
Yeah I 100% agree, no I'm not gonna go super much indepth with that right now
I also generate content and publish but i have two different results in two different websites.

This one got traction asap it got published
View attachment 528798

This one didn't. I am still working on it.
View attachment 528800

It totally depends how good i am fulfilling users query.
Thats super interesting, thanks for sharing - hope you'll keep us posted on how they end up doing.
There's a lot of factors imo, some of them are not obvious at all (at least they werent to me)
 
This matches what I’ve seen too. AI content itself is not the problem. The problem is when the output has no structure, no unique angle, and no reason for Google to treat it differently from every other article on the same topic. The prompt chain matters because it controls the research layer before the writing layer. I’d rather have AI build from:

- SERP intent
- entity coverage
- real examples
- internal link targets
- product/service context
- a clear page purpose
- a human edit pass before publishing

The worst AI content is usually just “write me 1,500 words about X.” That creates generic filler. The better version is closer to a workflow: outline from SERP gaps, add first-hand context, include specific data/examples, then edit for usefulness. That is probably why two sites with similar stats can perform completely differently.
 
Nice man. Which AI LLM did you use to create your content?
I use a bunch of different ones, I've combined combined GPT, Claude and Perplexity - I also use braves AI search API for data collection for each article
 
can you share some of your prompts with us??
No sorry

Good job.
I'd be interested to know how traffic conversion into $$$ is going, and what monetization model you're using?

The sites make $8000+ in total every month from affiliate marketing, could probably be more if I was added more affiliate links, but I try to not spam - I'm quite sure that could be the downfall of the sites :)
 
The sites make $8000+ in total every month from affiliate marketing, could probably be more if I was added more affiliate links, but I try to not spam - I'm quite sure that could be the downfall of the sites :)
Nice result :) Thanks for sharing!
Yes, I agree about affiliate links. Less is better than more.
 
No sorry



The sites make $8000+ in total every month from affiliate marketing, could probably be more if I was added more affiliate links, but I try to not spam - I'm quite sure that could be the downfall of the sites :)
because im same situation and my ai content keeps removing from serp

if u dont want to dhare what exactly prompts ur using

then you may just give some clue about it
 
Back
Top