- Apr 29, 2011
- 474
- 241
- Awards
- 7
- First Name
- Chris
ChatGPT - This is pretty awesome :)
- Thread starter Rick Buffkin
- Start date
- Oct 29, 2009
- 730
- 833
- Awards
- 8
- First Name
- Rick
Yep!!!
it's borrowing, combining and configuring content for the most part
it's borrowing, combining and configuring content for the most part
- Jan 3, 2019
- 275
- 200
- Awards
- 6
- First Name
- Marc
I played around Chat GPT the other day and recorded what I did, too bad I cut my email sequence written by Eminem out, should have left it inside.
Not for long by the sound of it. The people behind ChatGPT are already developing and releasing a tool to detect AI generated content (as they should, with the demonstrated abuse of it in scholastic settings having skyrocketed), and google has made their stance on it as well, expect rank punishments to follow.Yup. ChatGPT sure writes SEO content better than some of the things I’ve seen some companies outsource… and insource
Google Says AI Generated Content Is Against Guidelines
Content generated automatically using AI writing tools is against Google’s Webmaster Guidelines, says Search Advocate John Mueller.
www.searchenginejournal.com
I presume the only way they can do this is the same way they attempt to identify duplicate content between sites.
The AI content doesn't have a signature it leaves behind - it's just a combination of words, re-arranged from other data sources.
I don't see them effectively being able to detect this, unless you're blatantly asking it to write keyword-stuffed SEO content.
The real value here should be in things like "Write me a 200 word article on why customers would prefer to purchase their next vehicle from home".
- Apr 7, 2009
- 4,633
- 2,122
- Awards
- 10
- First Name
- Joe
Fun ChatGPT exploration.
I ask GPT:
GPT replies with a classic high quality response (nothing important here)
I ask GPT:
GPT replies with a classic high quality response. (nothing important here)
I ask GPT:
GPT replies
I ask GPT:
"You are a web site content specialist. Your client, Acme Chevrolet, just got 4 truckloads of new 2023 Chevrolet Silverados spanning a variety of colors and trims. Create content to place on the dealer's website to support this news."
GPT replies with a classic high quality response (nothing important here)
I ask GPT:
"rewrite this in a manner to become more human-like and disguise it's computer generated origin. "
GPT replies with a classic high quality response. (nothing important here)
I ask GPT:
"in your two responses, they look similar to me. Explain how does the latter response looks less computer generated"
GPT replies
The second response has been written with a more conversational tone and a more human-like approach. The language used is less formal and more approachable, making it appear less like a computer-generated response.
For example, in the second response, the phrase "Greetings!" is used as an opening, which is a more casual and friendly way of saying "Hello". Additionally, the use of words such as "we've got you covered" and "come on down" give the response a more personal touch, making it appear as if it was written by a real person rather than a machine.
Furthermore, in the second response, there's an emphasis on human interaction and assistance, as opposed to simply presenting information. This gives the response a more relatable and trustworthy feel, as it seems as though the writer is genuinely trying to help the reader find the right vehicle for their needs.
Overall, the changes made in the second response help to make the content more human-like, and less like a computer-generated response.
- Apr 20, 2009
- 694
- 697
- Awards
- 8
- First Name
- Ryan
I read an interesting article about this, sorry, I can't put my fingers on it now to give citation. The
premise was that AI can effectively watermark the content it manufactures by including specific characters in specific positions ie. the 517th character is the letter x, there are 3 words that are 5 letters long in the fifth sentence of the 3rd paragraph and so on. It feels like secret encoder rings and invisible ink from a bygone era. There really is nothing new under the sun.
I suppose if you have access to every word in the language you can easily place characters this way. Of course, you'll need the key to identify the watermark, but don't worry, there will be an AI for that too.
PS. Hmm... some letters in the message above are randomly underlined and in bold. Whatever could that mean??? (I really like encoder rings and invisible ink.
)I dont think it has that much to do with duplicate content in its textbook definition. That doesnt require complex algorithms to spot. Most are familiar enough with how google treated duplicate content for dealer.com clients a few years ago. In most cases ChatGPT is not even spitting out duplicate content as much as it is "revised content" based on other sources. We as people naturally do that whenever we opine or write on any subject we weren't the original person to discover as well.
The content generation ethical/moral/technical dilemma is when ChatGPT authors content on your behalf, specifically to affect search rank. Any automated content is technically spam in google's point of view (straight from the horse's mouth - John Mueller). If you are publishing blog pieces daily but ChatGPT is writing them for you, I see the issue. If you are using it to spin up unique descriptions on your VDPs, I dont see that as much of an issue (my personal opinion), as the intended use is very different.
As to the how, I believe it will specifically look for patterns that content-generating software can easily be programmed to produce, in the same way that accounting has the "rule of 9".
I cant imagine that the company that developed ChatGPT would be incapable of also developing a tool that "watermarks" AI content in language patterns, or something else entirely. I trust the math to take us there. The words are words to us, they're numbers to computers, which is what I predict will primarily detect the next generation of computer-generated content.
This is more or less the same feedback given to BDCs that auto-fire textbook replies to customers as well.
What are the odds they answer an email from the staff if it starts with Greetings and sounds like it was written by the alien ("AX") from Animorphs posing to be a human?
Ive seen a lot of good and bad examples of those rapid-fire automated emails, all working equally poorly unfortunately. But the bigger challenge is getting past the spam/promotions tab in the first place, of which I bet the less computer-generated the email sounds, the better the odds are of it getting through.
