🤫 Everyone Needs To Know About This ChatGPT SEO Secret

Make Exceptional ChatGPT SEO Content Using 3.5 (FREE)
Sometimes, we spend so long looking at new techniques and seeing what works, that we forget what was actually working in the first place. This is a great way to use AI for SEO
I am guilty of this, because I love to see what’s currently working and putting my own spin on it.
Sometimes, however, it’s better to stick to the things you know, and the methods you create.
In this case, it turns out that a lot of my content created using my previous method of prepping ChatGPT actually works extremely well.
In this video, I wanted to go back to this method, and explain why I think it might be working over other methods of huge one prompting.
Prepping vs. Prompting
I can’t say WHY prepping seems to work better, but I can give a few theories.
It’s common knowledge that ChatGPT works better when it understands what it’s actually doing, and giving it a really complicated prompt no doubt confuses it at times.
The long and complicated prompts are simply that, too long and too complicated. It’s the same with a lot of AI, if it’s able to give something good from a simple prompt, then it means that the output is going to be better, because the task isn’t quite as difficult for it.
It breaks it down into smaller parts, and it’s not trying to squeeze text into a certain amount of characters remaining
It works with ChatGPT 3.5 as well.
Prepping
Prepping is where instead of giving one giant prompt to ChatGPT with a lot of complex instructions, we instead ask it one step at a time to complete various tasks in the prompt. At the end of each prompt, we add “do you understand” meaning that ChatGPT will respond with yes I understand, and then store the information, instead of immediately starting to write.
Here’s an example of a prepping prompt for ChatGPT SEO, I am using 3.5 in this video purely because some people don’t want to pay for ChatGPT 4 - But I do recommend using GPT 4 for this
#chatgpt #chatgpt3 #seo #ai #aiseo
chat.openai.co...
Join this channel to get access to perks:
/ @incomestreamsurfers

[@heartlights]

Structuring a single prompt correctly can greatly improve the response based on the same logic as this "prepping" approach. However, I like how your approach allows room to still think out what I'm asking chatGPT for while I'm feeding it information.

[@byronhorde5892]

Been using Claude recently. That's what I have been doing with it. Making sure it understands what I am asking it to do. That is, before it actually does it.

[@Incomestreamsurfers]

I should be using Claude soon

[@chrisregan8463]

mate, thanks for all the great videos lately, Im wondering, with our websites, how important are backlinks?

[@sam-bartlett]

Agreed... I've struggled with long prompts and chat gpt4... prepping is the way to go! Learned it from you ;-)

[@nowar9220]

Thanks brother, your time, effort and secrets are greatly appreciated! When i turn a decent profit that stems from seo, il be sure to send you a care package, its only a matter of time before i can get my websitrs fully optimized and see them ranking with these strategies Lets gooo!!!

[@christophelourenco9193]

Prepping is what I always do. I never ever publish anything that doesn't come from a well-trained conversation. Thanks for the video Hamish

[@davidmiranski5237]

Great vid thanks Hamish. Have you considered using custom instructions in chat gpt it lets you create a kind of avatar and specify the output. It may clean up the prompting process and possibly make it remember things better. Is that feature for plus users i dont seem to have it.

[@Incomestreamsurfers]

They haven’t given me access yet

[@xXBlackWorkXx]

Can I use prepping using the 3.5 API tho?

[@AberrantArt]

I also noticed the new responses to the same exact prompts are more "politically correct" and cookie cutter compared to early days of chat gpt

[@ollieonai]

The reason why teaching chatgpt what to write is better than prompting it, is because of the bitsizes it can take. It gets confused when you give it too many at once.

[@MattJonesYT]

What do you mean by bitsizes? Is this referring to bit quantization?

[@junglestar]

No that’s not how it works. That’s not why step by step prompting is different. Learn promoting and prompt engineering

[@bluebird3461]

give us a prompts that writes 100% human generated content,,,,,, I'll write for amazon affiliate content like "best 10 coffee maker 2023"

[@LostYogi]

4:40 it could be that the old stuff has had longer to rank...

[@retratosariel]

I agree, it may be another reasons too, like redundant content, or single promoting, but it’s hard to believe that gpt3 is better content than 4, specially writing with good custom instructions and pluggings, keymate seems to work very well without hallucinating.

[@Incomestreamsurfers]

I’ve looked fairly deeply into it - my recent content even after a bit isn’t ranking as well as I would have expected. You could be right though

[@LostYogi]

@@Incomestreamsurfers I only just switched to 4.0 two weeks ago, but will keep an eye on my 3.5 vs 4.0 data

[@jadens9569]

This is certainly the way to get the most out of 3.5. I still use this method to date.

[@roydowding8191]

I noticed chatgbt 3.5 says to your human like prompt part today I'm sorry, but I cannot fulfill your request to create content that is intended to deceive AI content detectors or to artificially manipulate SEO. If you have any genuine requests for creating valuable and informative content, I'd be more than happy to help. Please let me know how I can assist you in a responsible and ethical manner.

[@MeowMovieInsightExplainedHindi]

After chatgpt output Then give again command Can you put list table of content on top

[@jessecunninghamv]

Great video Hamish 🚀🚀

[@vitaverse1607]

OMG!!! you delete my comments!!

[@Incomestreamsurfers]

KZbin automatically deletes comments - I did not delete it

[@FutureXT]

Awesome video Hamish 👍🏽 It's the reason I subbed to your channel the 1st time you presented this method a while ago✔️

[@journeyeu]

I feel gpt 3.5 wrote better content but it sometimes hallucinations and does what it want whatever your prompt is.😂

[@jeremy816]

Yep, good video. Prepping is so necessary. I've noticed a lot of generic responses, but when prepping, it's a different story. 👍

[@MattJonesYT]

Prepping in this context is a type of prompt chaining and I do agree that prompt chains at least two links long tend to do better than a single prompt, the weaker the model the more true that is and with very weak models there is no way to solve certain problems without long chains. GPT4 has been in the news that it can't do logic as well as it used to because they have apparently gimped its ability to do simple logic because people were using it to point out embarrassing incongruencies in the politically correct talking points that people are required to repeat in order to participate in today's version of a civilization. So going forward it will probably the the case that private offline models that don't have to tow the line for political correctness and can simply be actually correct will begin to outperform public models.

[@affiliateme7414]

I've heard a lot of people say 3.5 is better than 4 at a lot of things

[@MarkRhodesPhD]

Wow! When I just now checked -- 62 views in the first 6 minutes!! You're really in demand, and I love what you do for us...

[@comeon1405]

I'm second bro

[@JayBeeProduction]

good one

[@JayBeeProduction]

i have genius idea which noone ever had ... you prepp gpt 4!

[@Incomestreamsurfers]

That’s what I suggest doing in this video - it’s Even better. But a lot of people asked me to do this for 3.5. It works even better on 4.

[@JayBeeProduction]

@@Incomestreamsurfers thanks for your content man, started SEO couple weeks ago only cuz I found your ways

[@bradburyrobinson]

I'd absolutely agree with you on the main point of this video. I've tried both upfront prompting and continuous prompting and I find the latter works slightly better, it definitely feels like the further you get into a chat, CGPT seems to 'forget' earlier instructions. I'm not sure if this is a technical limitation of these models, something to do with the token parameters or that. It's annoying when you have clearly established something but then it goes off track. Secondly, your point about telling it to use lists and tables. I've also had this, I told it to include more lists and it went nuts and there were lists inside lists with lists for content, etc. I also once told it to be a bit more casual in its writing style and it turned into Danny Dyer down the pub after ten pints. I had to knock that one on the head quickly.

[@cestmoi8372]

Ha ha I've found this. I once asked it to make sure it included UK English spellings and it came out with something virtually Shakespearian 😂