- Joined
- Sep 3, 2014
- Messages
- 6,244
- Likes
- 13,130
- Degree
- 9
I just saw easily one of the smartest SEO moves I've ever encountered. Here's the video which is 6 minutes long (so 3 mins on 2x speed) that I'll also break down below:
So the backstory is people on 4chan realized they could get ChatGPT to not only get around it's inbuilt biases but even expose them by having it create a table ranking political figures between -10 and +10, and it goes exactly how you'd think.
Fast forward and I'm assuming Matt Diggity or some SEO saw this too and it got their brain churning.
Diggity uses the Product Reviews guidelines for his example but it would work for Helpful Content or anything else. He pastes in the guidelines into ChatGPT and asks it to distill it down to the 5 most important points.
From there, he then pastes in a product review and asks ChatGPT to evaluate the article based on those 5 criteria and rate them between 0 and 10. He suggests that anything at a 7 or below be improved.
So for example, the article he pastes in has plenty of "show the reviewer using the product" but it's all in a video which Google can't understand.
At this point he asks ChatGPT to write a piece of content that helps improve that score that you can rewrite or not and plop right under the video to improve this part of the score.
He then suggests creating a spreadsheet of all your posts, with columns for each of the 5 criteria, and drop the scores in. If the article is too long to paste in, he says break it into halves and get the scores and use the highest score for each criteria. Do this and collect all your data, and let the spreadsheet highlight anything at a 7 or below.
Then you set about going through the process of improving the low scores.
The fail point of this is assuming ChatGPT can understand the guidelines good enough to distill it down to the 5 most important points. Based on how much time you have you could do 3 points, 5 points, 7 points, etc. The 2nd fail point is assuming ChatGPT can understand your content well enough and spit out something valuable to add to the article. The issue being if you have 1,000 posts it'll take a long time to do this "right", so you'll have to trust the bot on it.
I have to hand it to whoever thought this up. This is genius-level thinking. I'm super impressed.
So the backstory is people on 4chan realized they could get ChatGPT to not only get around it's inbuilt biases but even expose them by having it create a table ranking political figures between -10 and +10, and it goes exactly how you'd think.
Fast forward and I'm assuming Matt Diggity or some SEO saw this too and it got their brain churning.
Diggity uses the Product Reviews guidelines for his example but it would work for Helpful Content or anything else. He pastes in the guidelines into ChatGPT and asks it to distill it down to the 5 most important points.
From there, he then pastes in a product review and asks ChatGPT to evaluate the article based on those 5 criteria and rate them between 0 and 10. He suggests that anything at a 7 or below be improved.
So for example, the article he pastes in has plenty of "show the reviewer using the product" but it's all in a video which Google can't understand.
At this point he asks ChatGPT to write a piece of content that helps improve that score that you can rewrite or not and plop right under the video to improve this part of the score.
He then suggests creating a spreadsheet of all your posts, with columns for each of the 5 criteria, and drop the scores in. If the article is too long to paste in, he says break it into halves and get the scores and use the highest score for each criteria. Do this and collect all your data, and let the spreadsheet highlight anything at a 7 or below.
Then you set about going through the process of improving the low scores.
The fail point of this is assuming ChatGPT can understand the guidelines good enough to distill it down to the 5 most important points. Based on how much time you have you could do 3 points, 5 points, 7 points, etc. The 2nd fail point is assuming ChatGPT can understand your content well enough and spit out something valuable to add to the article. The issue being if you have 1,000 posts it'll take a long time to do this "right", so you'll have to trust the bot on it.
I have to hand it to whoever thought this up. This is genius-level thinking. I'm super impressed.