[Journey] Experimental Blog using ChatGPT

Just to clarify something.
And this is gonna sound like a dumb question probably.

But have you really built a site with just AI content, no backlinks, and 3 months later doing over 100 PV a day and $2 in revenue?

I kind of feel like I've missed something.
There are no dumb questions. Yes, I guess I was lucky with this one. I have not been able to work on this site this month as I've had some problems with my main project. But the traffic now sits at around 200 organic visits per day and about 65 articles in total.

The site is 100% AI, no keyword research, no backlinks, absolutely no competition, and for some weird reason a lot of searches. I use one prompt to generate the entire article and post it without any human edit at all. Each post is 600-800 words and I only grab a review photo from Amazon to use as thumbnail.

A few examples would be: A site about weights. How much does X weigh, converted into both metric and imperial units. You don't need to do any research in such a niche, just grab a list (or ask chatgpt to make such a list) of different items, foods, products, etc. Another one could be weight loss related and only focus on carbs in different types of foods. Or the pH value of different liquids. Of course you have to fact check the article to make sure you're not spreading false statistics.
 
There are no dumb questions. Yes, I guess I was lucky with this one. I have not been able to work on this site this month as I've had some problems with my main project. But the traffic now sits at around 200 organic visits per day and about 65 articles in total.

The site is 100% AI, no keyword research, no backlinks, absolutely no competition, and for some weird reason a lot of searches. I use one prompt to generate the entire article and post it without any human edit at all. Each post is 600-800 words and I only grab a review photo from Amazon to use as thumbnail.

A few examples would be: A site about weights. How much does X weigh, converted into both metric and imperial units. You don't need to do any research in such a niche, just grab a list (or ask chatgpt to make such a list) of different items, foods, products, etc. Another one could be weight loss related and only focus on carbs in different types of foods. Or the pH value of different liquids. Of course you have to fact check the article to make sure you're not spreading false statistics.
Well done.
This is awesome and I'm glad you're doing so well.
I have a few sites on the go at the moment, one is 3 months old but only about 5 posts. It was originally setup for a business I was assisting with.
Haven't done anything to it and it's not even indexed.
Another site I setup about 3 weeks ago and added some content to last week already has every post indexed.
So its a bit hit and miss.
No back links created or any real keyword research done on either. And no requests to index via gsc.

It's interesting to see how things seem to randomly work while others don't.
 
Good luck mate :)
Just another interesting journey based on AI articles.
However, my experience has not been good with AI articles but I'm still trying my luck by implementing different strategies to make it work. I'm doing my experiments until I find the best prompt.
 
Good luck mate :)
Just another interesting journey based on AI articles.
However, my experience has not been good with AI articles but I'm still trying my luck by implementing different strategies to make it work. I'm doing my experiments until I find the best prompt.
Thanks!

I think there are a lot of factors to consider when doing AI websites. For instance, chatgpt is terrible with affiliate reviews, it completely lacks the human connection and it won't convert at all (I'm actually making a 100% affiliate site on the side just to try and see how it will perform). I find that short info posts, in niches with no competition is the best way to go. That's what I did with this one and I'm preparing one more site in a completely different niche but it's a good one too.

The niches I find work the best are the same niches that would suit a programmatic SEO approach. Where you have one key question that can apply to multiple products/states/countries/animals, etc. Start with something simple and see what works for you. There's a guy on YouTube that does massive AI sites about birds in different languages, he's doing really well it seems. Might want to try your hand at some medium sized language like German or French.
 
July Update

Another month has passed and the website is still in the game. I have had some problems with Ezoic on both of my journeys, but this one was not as bad. However, I still decided to take this site off Ezoic for the time being until I figure everything out.

The site now has 80 live articles, way less than what I was hoping. There are another 15 or so articles in draft, just gotta edit them and add thumbnails.

Let's get into the stats for the month:

SERPRobot

serprobot-july.jpg

GSC

gsc-mar-july.jpg

Analytics

analytics-july.jpg

I still don't like Google Analytics 4, but whatever. As you can see I had a huge dent in traffic in late July. This is not because Google caught up with my AI content, when I changed nameservers from Ezoic it took way longer than I was expecting. The site was down for 30 hours or more stretching over 2 days.

The site is running smoothly now, and traffic seems to be growing nicely. Probably going to get back on Ezoic soon, as AdSense pays pennies in this niche...

The number of unique visitors to the site in July was 4.1k and we've passed 200 daily visitors in a single day.

Earnings

July was the first month this site ran ads, and we did earn $24.09 with an average ePMV of $6.18. Considering the site is 4 months old and we only showed ads between the 7th and 28th of July, that's pretty decent. I really want to get this origin error sorted and get back on the horse.

What's the plan going forward?

Now that I've seen the site grow month after month, I really want to see its true potential. I believe it's possible to grow to 50k+ monthly visitors and get on Mediavine, there's no doubt in my mind about it. It might take a few more months though (or a year), but I'm not in a hurry.

As I'm having some issues with my other site, I have to put more attention towards fixing that. Going on vacation in August, traveling around Europe so won't have a lot of time. However, I will try to front-load some scheduled articles and grow the site to 100+ published posts before the next update.

I have a few articles with great potential, they currently rank #10 but I do know for a fact that they bring in 6k+ visitors per month for the guy ranking #1. His site is weak but mine is weaker. Thinking of possibly getting backlinks to these posts but I'm just not ready to invest in this site yet. We'll see how the articles rank in the next update.
 
Good Luck! Are you using complex chatgpt prompts for these posts? Many people using chatgpt these days get some traffic in the beginning, but eventually get deindexed, probably because they aren't using good enough prompts.
I am also using chatgpt but with simple prompts. Can you please give advance prompt example.

If a keyword is weight loss product review "x" then how will you write the prompt.

Thanks
 
Thanks!

I think there are a lot of factors to consider when doing AI websites. For instance, chatgpt is terrible with affiliate reviews, it completely lacks the human connection and it won't convert at all (I'm actually making a 100% affiliate site on the side just to try and see how it will perform). I find that short info posts, in niches with no competition is the best way to go. That's what I did with this one and I'm preparing one more site in a completely different niche but it's a good one too.

The niches I find work the best are the same niches that would suit a programmatic SEO approach. Where you have one key question that can apply to multiple products/states/countries/animals, etc. Start with something simple and see what works for you. There's a guy on YouTube that does massive AI sites about birds in different languages, he's doing really well it seems. Might want to try your hand at some medium sized language like German or French.
Are you talking about julian of chipperbirds :)
 
Are you talking about julian of chipperbirds :)
Yes, that's the guy. I know we are doing two completely different strategies, he's just blasting thousands of posts and still manages to rank. I have not been able to pull that off.

I am also using chatgpt but with simple prompts. Can you please give advance prompt example.

If a keyword is weight loss product review "x" then how will you write the prompt.

Thanks
I will try to answer this question quickly, however, I don't do affiliate posts so will use an informational post as an example. Some people make one prompt per section of the article, I just do a one-shot prompt and that's good enough. The most important thing is that you have to be strict, tell it exactly what you want.

Obviously, I don't write about airplanes, this is just a quick example.

This is typically how my prompt starts:

Code:
I have a website about airplanes, specifically sailplanes.
I want you to write a detailed article with the following title: "What's The Difference Between a Glider and a Motor Glider?".
I want you to mix long and short sentences, and make it not detectable by AI detectors, but don't use uncommon words or phrases.

I have seen good results just telling ChatGPT to make it non-detectable. I also tell it to exclude specific words, therefore, moreover, additionally, however, in conclusion, are a few words that make it look more AI written. It won't completely ignore these types of words but it uses them way less often.

The next part is something like this:

Code:
I want you to use the following outline with H2 and H3 headings, paragraphs, bulleted lists, and links to sources where applicable:

2 introduction paragraphs
H2: Difference between a Glider and a Motor Glider
2 paragraph
H3: Motor
2 paragraphs
H3: Speed
2 paragraphs
H3: Altitude
2 paragraphs
H3: Cost
2 paragraphs
H2: Pros and Cons of a Glider
1 paragraph
1 list of 5 pros
1 list of 3 cons
H2: Pros and Cons of a Motor Glider
1 paragraph
1 list of 5 pros
1 list of 3 cons
H2: Should You Get a Glider or a Motor Glider?
2 paragraphs
H2: Conclusion
1 paragraph

If you don't write out exactly how many paragraphs, lists etc you want for each heading separately, it will just write at random. Most of the times you will end up with an article of 700 words. If you do this however, you can build a 3000-word article in one prompt easily. 3000 words is a lot to fact-check and edit, so I tend to keep the outline pretty short anyways, but 1000-1500 words is good.

If you're writing an article about a specific topic or product, you can copy and paste the exact specifications from another site into the start of the prompt. This will help make the article more correct.
 
Yes, that's the guy. I know we are doing two completely different strategies, he's just blasting thousands of posts and still manages to rank. I have not been able to pull that off.


I will try to answer this question quickly, however, I don't do affiliate posts so will use an informational post as an example. Some people make one prompt per section of the article, I just do a one-shot prompt and that's good enough. The most important thing is that you have to be strict, tell it exactly what you want.

Obviously, I don't write about airplanes, this is just a quick example.

This is typically how my prompt starts:

Code:
I have a website about airplanes, specifically sailplanes.
I want you to write a detailed article with the following title: "What's The Difference Between a Glider and a Motor Glider?".
I want you to mix long and short sentences, and make it not detectable by AI detectors, but don't use uncommon words or phrases.

I have seen good results just telling ChatGPT to make it non-detectable. I also tell it to exclude specific words, therefore, moreover, additionally, however, in conclusion, are a few words that make it look more AI written. It won't completely ignore these types of words but it uses them way less often.

The next part is something like this:

Code:
I want you to use the following outline with H2 and H3 headings, paragraphs, bulleted lists, and links to sources where applicable:

2 introduction paragraphs
H2: Difference between a Glider and a Motor Glider
2 paragraph
H3: Motor
2 paragraphs
H3: Speed
2 paragraphs
H3: Altitude
2 paragraphs
H3: Cost
2 paragraphs
H2: Pros and Cons of a Glider
1 paragraph
1 list of 5 pros
1 list of 3 cons
H2: Pros and Cons of a Motor Glider
1 paragraph
1 list of 5 pros
1 list of 3 cons
H2: Should You Get a Glider or a Motor Glider?
2 paragraphs
H2: Conclusion
1 paragraph

If you don't write out exactly how many paragraphs, lists etc you want for each heading separately, it will just write at random. Most of the times you will end up with an article of 700 words. If you do this however, you can build a 3000-word article in one prompt easily. 3000 words is a lot to fact-check and edit, so I tend to keep the outline pretty short anyways, but 1000-1500 words is good.

If you're writing an article about a specific topic or product, you can copy and paste the exact specifications from another site into the start of the prompt. This will help make the article more correct.
This is useful . I am not using chatgpt in detail like you. I will try to follow your format for my posts. Also check competitors sites how they are formatting the articles. I appreciate your work. Smile
 
This is useful . I am not using chatgpt in detail like you. I will try to follow your format for my posts. Also check competitors sites how they are formatting the articles. I appreciate your work. Smile
Yes but keep in mind you have to experiment what works best for your style. Have a look at the competition, don't just include the same headings as your competitors in the prompt, make them better.
 
It seems your AI site is doing better compared to what your main site did in the same time span :)
Best of luck mate.
 
It seems your AI site is doing better compared to what your main site did in the same time span :)
Best of luck mate.
That's right! But I still want to focus on building that up as I plan to sell it. Would be interesting to see if it's possible to build a portfolio of AI sites and make a living from that.

I already have this one and then I'm experimenting with an affiliate related AI site. Looking into creating one more in a niche I'm hyper interested in. Also want to do some experiments with Pinterest. So even if this AI site would go down I would probably just keep updating this journey with new experiments.
 
What do you think how much AI can go deep with some techical things in articles?
Ofc in order to warm the reader's attention with some valauble things he cand read on site.
 
What do you think how much AI can go deep with some techical things in articles?
Ofc in order to warm the reader's attention with some valauble things he cand read on site.
I don't think you should use AI for very technical posts, it just won't turn out very well. It's better to focus on simple things, and information that you easily can fact check before publishing.
 
Yes, that's the guy. I know we are doing two completely different strategies, he's just blasting thousands of posts and still manages to rank. I have not been able to pull that off.


I will try to answer this question quickly, however, I don't do affiliate posts so will use an informational post as an example. Some people make one prompt per section of the article, I just do a one-shot prompt and that's good enough. The most important thing is that you have to be strict, tell it exactly what you want.

Obviously, I don't write about airplanes, this is just a quick example.

This is typically how my prompt starts:

Code:
I have a website about airplanes, specifically sailplanes.
I want you to write a detailed article with the following title: "What's The Difference Between a Glider and a Motor Glider?".
I want you to mix long and short sentences, and make it not detectable by AI detectors, but don't use uncommon words or phrases.

I have seen good results just telling ChatGPT to make it non-detectable. I also tell it to exclude specific words, therefore, moreover, additionally, however, in conclusion, are a few words that make it look more AI written. It won't completely ignore these types of words but it uses them way less often.

The next part is something like this:

Code:
I want you to use the following outline with H2 and H3 headings, paragraphs, bulleted lists, and links to sources where applicable:

2 introduction paragraphs
H2: Difference between a Glider and a Motor Glider
2 paragraph
H3: Motor
2 paragraphs
H3: Speed
2 paragraphs
H3: Altitude
2 paragraphs
H3: Cost
2 paragraphs
H2: Pros and Cons of a Glider
1 paragraph
1 list of 5 pros
1 list of 3 cons
H2: Pros and Cons of a Motor Glider
1 paragraph
1 list of 5 pros
1 list of 3 cons
H2: Should You Get a Glider or a Motor Glider?
2 paragraphs
H2: Conclusion
1 paragraph

If you don't write out exactly how many paragraphs, lists etc you want for each heading separately, it will just write at random. Most of the times you will end up with an article of 700 words. If you do this however, you can build a 3000-word article in one prompt easily. 3000 words is a lot to fact-check and edit, so I tend to keep the outline pretty short anyways, but 1000-1500 words is good.

If you're writing an article about a specific topic or product, you can copy and paste the exact specifications from another site into the start of the prompt. This will help make the article more correct.
wow, I thought ChatGPT must have had a limit to how much it could output because I could only ever get 400-800 words out of it. Even when I asked for a 1500 word article.


But I was giving it very simple instructions.


Eg using your example of airplanes.

I'd do a very quick look at a keyword I wanted to target, maybe How likely is a propeller to break?
I'd then ask ChatGPT to come up with an epic Blog title for that.
Then I'd ask it to write me an epic and engaging blog post as long as possible while using facts about (insert the title it gave me)
And that would be about it.
A lot of editing required though with the final output.

Another I'd do for a bigger post.
I'd ask ChatGPT for 20 of the most common pain points for airplane owners
Then I'd use the same above prompt for each of the 20 things that chatgpt told me about.
I could never ask it for all 20 at once because it would only give me very short info about each and keep the total word count well under 1,000 when realistically you'd want at least 100-200 words for each of the 20.

I must admit though, I only ever did VERY minimal reading into what are some great ChatGPT prompts and tried a few out and went with the first that semi-worked.
 
wow, I thought ChatGPT must have had a limit to how much it could output because I could only ever get 400-800 words out of it. Even when I asked for a 1500 word article.


But I was giving it very simple instructions.


Eg using your example of airplanes.

I'd do a very quick look at a keyword I wanted to target, maybe How likely is a propeller to break?
I'd then ask ChatGPT to come up with an epic Blog title for that.
Then I'd ask it to write me an epic and engaging blog post as long as possible while using facts about (insert the title it gave me)
And that would be about it.
A lot of editing required though with the final output.

Another I'd do for a bigger post.
I'd ask ChatGPT for 20 of the most common pain points for airplane owners
Then I'd use the same above prompt for each of the 20 things that chatgpt told me about.
I could never ask it for all 20 at once because it would only give me very short info about each and keep the total word count well under 1,000 when realistically you'd want at least 100-200 words for each of the 20.

I must admit though, I only ever did VERY minimal reading into what are some great ChatGPT prompts and tried a few out and went with the first that semi-worked.
Yeah I get that sometimes too and have to type continue for it to keep writing. The thing is, ChatGPT knows nothing about titles and keywords. You should always do keyword research yourself, then mimick the competition in terms of the outline.
 
I've just gone through my rankings as well as GSC. There are some articles that rank #6-10 with 10-20 clicks in about a months time, could be huge traffic if I get them to rank #1. There are also a couple of articles that I know for a fact I won't get any traffic from, they rank #1 and have had 1-2 clicks for the past 3 months.

I'm considering removing those articles, as I would like to have as few, but high traffic articles as possible. My thoughts is that having hundreds of AI articles could be risky if I'm trying to pass the site off as legit. Reducing the number of articles, while still keeping the best performers, I should have an easier time flying under the radar. Besides, if I decide to change the layout and style of the blog posts it should be a lot quicker.

The few articles that has the most potential might need a few backlinks to rank #1. These are not really following my original pattern of posts so they are unfortunately not 0 competition like the others.

Anyone got any experience removing non-performing blog posts? I guess I should 301 them to the homepage or related posts.
 
Back
Top