How do you check if content was generated by ChatGPT (OpenAI GPT-3/3.5/4)?

[just a punk]

I've tried these:

https://x.writefull.com/gpt-detector
the results are about this:
6% likely this comes from GPT-3, GPT-4 or ChatGPT.

https://writer.com/ai-content-detector/
the results are about this:
97%
HUMAN-GENERATED CONTENT
Fantastic!

https://gptzero.me/
the results are about this:
Your text is likely to be written entirely by a human

https://www.zerogpt.com/
the results are about this:
Your Text is Human written
0%
AI GPT*

The problem is that I've been trying to check the content generated by OpenAI GPT 3.5 Turbo for my websites. It is not human-written for sure, so all those results are absolutely wrong.

Is there any online service that is really able to detect a GPT generated text? Is there something that actually works?

[Rochard]

LOL. If it says "In conclusion" it it most likely written by Chat GPT.

[Klen]

I use zerogpt too, however it does detect most of AI content as AI but not everything so you just happen to get negative result.

[just a punk]

Originally Posted by Rochard

LOL. If it says "In conclusion" it it most likely written by Chat GPT.

With a 6% probability? Nice point, Sherlock

Originally Posted by Klen

I use zerogpt too, however it does detect most of AI content as AI but not everything so you just happen to get negative result.

All posts give about the same result. I just ask GPT to write "as a human" and use my plugin's built-in synonymizer, which sometimes replaces "that" and "this" with "the". That's all I actually do, and none of these online checkers are able to detect GPT there...

Maybe OpenAI has changed something on their side and improved the quality of the generated content. Anyway, it seems that all these GPT checkers are just useless. I'd like to find a working one to play with. Any suggestions?

P.S. That was checked with Turbo GPT-3.5 model only. As a developer, I have an access to GPT-4 model, but it's too expensive and very slow. So ATM GPT-3.5 is a my favorite.

[DrGentle]

Originally Posted by CyberSEO

Anyway, it seems that all these GPT checkers are just useless.

I 100% agree with this.

[Klen]

Originally Posted by CyberSEO

With a 6% probability? Nice point, Sherlock



All posts give about the same result. I just ask GPT to write "as a human" and use my plugin's built-in synonymizer, which sometimes replaces "that" and "this" with "the". That's all I actually do, and none of these online checkers are able to detect GPT there...

Maybe OpenAI has changed something on their side and improved the quality of the generated content. Anyway, it seems that all these GPT checkers are just useless. I'd like to find a working one to play with. Any suggestions?

P.S. That was checked with Turbo GPT-3.5 model only. As a developer, I have an access to GPT-4 model, but it's too expensive and very slow. So ATM GPT-3.5 is a my favorite.

You should be happy if it passes as human content lol. Furthermore, I am confident that in the future, AI content will go completely undetected. And also what i noticed, how content that states facts, for example, if you prompt "generate an article about historical facts," will pass as human content.

[just a punk]

I also think OpenAI will improve their models to be undetectable. Even now, a proper prompt might pass all those online GPT checkers. Personally, I'd like to get access to the one used by Google, but of course that's an impossible dream.

Anyway, a simple built-in synchronizer allows anyone to cheat any online checker with ease. DeepL translation of a GPT-generated text and/or the old text spinners from 2010's also work just perfectly. All these methods can be used in a fully automated way, so autoblogging is alive again and now I don't see anything that can stop it.

Human blog writers and obsolete. Period.

[TheLegacy]

Sad really when you think that a well drafted article by a human with a good command of the English language suddenly can be accused of AI. It's one of the lost arts of our times that was lost in the school system.

That being said - I haven't found an accurate system that can be relied upon to spot AI generated text - the only way is really knowing the author and ask, "are they really that smart?"

[Klen]

google just announced how they will publish tools to recognize AI content.

[CaptainHowdy]

Originally Posted by CyberSEO

Human blog writers and obsolete. Period.

More free time to work on that quirky novel, guys . . .

[just a punk]

Originally Posted by Klen

google just announced how they will publish tools to recognize AI content.

They'll be useless as well

[elron]

aren't those cheap versions of the Alan turing test from 80 years ago?

[emmasexytime]

Originally Posted by Rochard

LOL. If it says "In conclusion" it it most likely written by Chat GPT.

[Klen]

I did another test with following prompt:
write a description for meal with following ingredients : Baked chicken gyros in Mexican sauce with cheese, American potatoes

First i did chat gpt 3.5 - detected as 100%
Then did bing chat which is based on gpt 4 - detected as 66%
And finally i changed prompt to "write in style of food critic" then it went to zero.

[TheLegacy]

Originally Posted by Klen

I did another test with following prompt:
write a description for meal with following ingredients : Baked chicken gyros in Mexican sauce with cheese, American potatoes

First i did chat gpt 3.5 - detected as 100%
Then did bing chat which is based on gpt 4 - detected as 66%
And finally i changed prompt to "write in style of food critic" then it went to zero.

Curious - I don't think readers will care and it does not hurt you SEO wise by Google. In a recent post, Google confirmed that AI-generated content is not against its search engine guidelines

https://developers.google.com/search...and-ai-content

[Klen]

Originally Posted by TheLegacy

Curious - I don't think readers will care and it does not hurt you SEO wise by Google. In a recent post, Google confirmed that AI-generated content is not against its search engine guidelines

https://developers.google.com/search...and-ai-content

There is FB group called "Chatgpt and SEO" and people inside claim how google recognize AI content but bing does not so you still get traffic.

[TheLegacy]

Originally Posted by Klen

There is FB group called "Chatgpt and SEO" and people inside claim how google recognize AI content but bing does not so you still get traffic.

Search Engine Market Share: According to SimilarWeb

https://www.similarweb.com/engines/

As of April 2023

Google 91.05%
Bing 2.92%
Yahoo 2.87%
Yandex .61%

So - Yes Bing may not but is that really going to hurt your SEO if you use Chatgpt?

[twinklover]

There are things you can do to get the output to bypass detection. If you tell the tool to use certain vocabulary it may do it. But it's possible that you don't need to bother.

[Klen]

There is another detector but it cost , originality.ai and turnitin. Plus one more tip how to humanize your prompts - tell it to start sentence with words like "i did" , "i loved it" , "i enjoyed it" and voila no ai detector will be ever able to figure out.

[just a punk]

Originally Posted by Klen

There is FB group called "Chatgpt and SEO" and people inside claim how google recognize AI content but bing does not so you still get traffic.

It really depends on how you use it. If you use it for "evil" like rewriting news or even promoting products from online markets (here I would suggest to use a synonymizer on your GPT-generated content), Google will really cut off your traffic. If you use it for "good" to generate unique content of your own specific topic, Google is okay with it.

I have rewritten the CyberSEO Pro manuals and main site pages with ChatGPT and... its positions in Google SERPs have instantly... improved

Check this out: https://www.cyberseo.net/blog/should...rated-content/