Chat-GPT "auto-content" limitations

Hey all,
First pro-active post and hoping to add some value to the community (have taken tons).

Thought it was worth sharing my experience when it came to using chat-GPT to create content at scale.

I generated ~2000 articles without too much trouble, with roughly 25 articles per niche, with the intention being to see what (if anything) stuck with on page SEO only. Short answer is not a lot

But the more troubling answer, is where I did generate interest / search impressions and took a look at the content, it was - to be blunt - complete bollocks. Not an exact e.g. but a GPT list of "vegetarian only restaurants in new york" would include multiple steak houses. Worth considering that this is info that fits various google local data fields, so it's not "seeing" those.

Basically: it can create plausible word salad but it can't be trusted at all for info quality, if that's a priority.

I've noticed two trends from YouTubers, and AI.

One AI is unreliable on quality consistently, especially bard it will throw a mix salad and you'll be greatul at first but frustrated later on, that's why it's called a tool until regulated.

The second one is YouTubers using AI sites to promote their audience(that's fine and all)until the service removes the ability to make multiple accounts with VPN blocks because of the sudden notice.


However you look at it, you'll always have to do a bit of the dirty work eventually or pass it onto a virtual assistant.

Factual correctness and staying aware of the context is the biggest challenge for GPT for sure.

But maybe the steak houses it picked have veggie only menus?

Gants-P said:

I've noticed two trends from YouTubers, and AI.

One AI is unreliable on quality consistently, especially bard it will throw a mix salad and you'll be greatul at first but frustrated later on, that's why it's called a tool until regulated.

The second one is YouTubers using AI sites to promote their audience(that's fine and all)until the service removes the ability to make multiple accounts with VPN blocks because of the sudden notice.


However you look at it, you'll always have to do a bit of the dirty work eventually or pass it onto a virtual assistant.

Click to expand...

In essence, I'm tending to think of AI as a very low grade virtual assistant. Give extremely specific instructions and you'll probably get an "ok" result, but will need to validate that work.

tiiberius said:

Factual correctness and staying aware of the context is the biggest challenge for GPT for sure.

But maybe the steak houses it picked have veggie only menu!

Click to expand...

Imperfect analogy.... It's like I asked for "kid friendly" and they returned places listing "no kids".

My kids will eat what their told, so steak or otherwise is neither here nor there... ;-)

Ballan said:

In essence, I'm tending to think of AI as a very low grade virtual assistant. Give extremely specific instructions and you'll probably get an "ok" result, but will need to validate that work.

Click to expand...

Yup, that's why experts in the industry that's not in threat of losing their jobs due to repetition have said AI is just a tool a well automated tool that will just get worse when it mutates with other AI open source models.

Facebook's lama model breeding with open AI's model with multiple breeds of mutations.


Yup definitely gonna get worse and only Microsoft will stand tall building trust for who can use their AI.

Ballan said:

Hey all,
First pro-active post and hoping to add some value to the community (have taken tons).

Thought it was worth sharing my experience when it came to using chat-GPT to create content at scale.

Click to expand...

As far as I can see, we cannot really consider what you describe as a real "experience" using "chat-GPT"...

Ballan said:

I generated ~2000 articles without too much trouble, with roughly 25 articles per niche, with the intention being to see what (if anything) stuck with on page SEO only. Short answer is not a lot

Click to expand...

All DEPENDS on HOW you did manage to get "generated 2000 articles" and this is exactly where the 'problem' starts. Especially if you've been brought to believe all the hype and myths (or instant magic) about AI.

Ballan said:

But the more troubling answer, is where I did generate interest / search impressions and took a look at the content, it was - to be blunt - complete bollocks. Not an exact e.g. but a GPT list of "vegetarian only restaurants in new york" would include multiple steak houses. Worth considering that this is info that fits various google local data fields, so it's not "seeing" those.

Click to expand...

See, we pointed the 'problem' already. And it's what you ignore about AI versus what you've been brought to believe.
AI is a tool, not a magic wand.

Ballan said:

Basically: it can create plausible word salad but it can't be trusted at all for info quality, if that's a priority.

Click to expand...

What you expose here, unfortunately, and well fortunately at the same time as it helps better understand a little more, is exactly one the unresponsible ways to use it. Moreover, it can put you in SERIOUS HOT WATER, even unwittingly...