WARNING! Don't copy and paste texts from generators like chatgpt!

[jyciap11]

Hi everyone, I want to give some advice to anyone who creates websites, articles or other things with artificial intelligence (Claude, ChatGPT, etc.) with the aim of being successful on Google.
Do NOT copy and paste text from chatgpt directly into wordpress, but format it first!!
I have seen dozens of sites that were indexed, receiving traffic, and then suddenly being brutally killed by Google. I couldn't understand why until I realized that the problem was copy and paste. Copying and pasting generates strange source codes (take a look at the code I've attached) and google uses these to understand that it's text copied and pasted by AI generators.
Here are some examples:

<h3 data-pm-slice="0 0 []">

<h1 class="font-bold text-h2 leading-[36px] pt-[30px] pb-[7px] [&amp;_a]:underline-offset-[6px] [&amp;_.underline]:underline-offset-[11px]" dir="ltr" style="text-align: center;">

<h2 class="font-bold text-h3 leading-[31px] pt-[11px] pb-[5px] [&amp;_a]:underline-offset-[6px] [&amp;_.underline]:underline-offset-[2px]" dir="ltr">

Always format the text before using it, it cost me time and money before I understood!

 

[xReminisce]

Yes, AI watermarks them
thought that was common knowledge..
well it is now lol

 

[Spicy Chromosomes]

Are you manually formating those texts?

 

[jyciap11]

Are you manually formating those texts?

Nope, I copied and past for months LOL, without understand why my websites got penalized

 

[hideath]

You can specifically ask for the HTML code of the output, mention the required and possibly used HTML elements,
The same applies to asking for the content to be output as Markdown, which could be easily copy-pasted and automatically formatted with WP,

The output HTML attributes don't look like any kind of watermarking either, it's just part of their Frontend CSS code,
You mostly got doomed because of the content output itself or other SEO-related stuff

 

[tiiberius]

Yes.

Unfortunately, what is also a footprint is the text itself.

I suspect search engines have sampled (we're talking millions of samples) what popular GPTs produce and trained their own AI to detect similarities.

For example:
Both search engines and users prompt: "Write an article about [keyword]". Resulting text is fairly similar.

Now if you copy and paste, they can easily see how much it matches their sample.

 

[jyciap11]

You can specifically ask for the HTML code of the output, mention the required and possibly used HTML elements,
The same applies to asking for the content to be output as Markdown, which could be easily copy-pasted and automatically formatted with WP,

The output HTML attributes don't look like any kind of watermarking either, it's just part of their Frontend CSS code,
You mostly got doomed because of the content output itself or other SEO-related stuff

I'm not really sure about this. The content is really HQ, but if the code has "[]" in many parts of the code.. I don't think it's fine at Google eyes.
"[]", for me, it's a watermark, or in any case it's something suspicious for Google.
Btw also "leading-[36px] pt-[30px] pb-[7px] [&_a]" it's not clear content, seems something generated at the google eyes for me.
I have many wordpress sites written myself, and the code it's clear, without shit like "[]"

 

[jyciap11]

Found on X other people that ask about this "strange" watermark.

https://twitter.com/x/status/1859569190861951376

Really interesting about this, why chatgpt add this to the code, instead of something like "generated by chatgpt etc. etc."?

 

[FanTechstic]

Interesting, thanks for sharing

 

[hideath]

Found on X other people that ask about this "strange" watermark.

https://twitter.com/x/status/1859569190861951376

Really interesting about this, why chatgpt add this to the code, instead of something like "generated by chatgpt etc. etc."?

I'm not really sure about this. The content is really HQ, but if the code has "[]" in many parts of the code.. I don't think it's fine at Google eyes.
"[]", for me, it's a watermark, or in any case it's something suspicious for Google.
Btw also "leading-[36px] pt-[30px] pb-[7px] [&_a]" it's not clear content, seems something generated at the google eyes for me.
I have many wordpress sites written myself, and the code it's clear, without shit like "[]"

Just a little bit of digging and here is what they're used for:
They're part of the "Tailwind CSS" CSS-Framework: https://tailwindcss.com/,
You can find similar codes in their docs here: https://tailwindcss.com/docs/line-height

If we're assuming your statement is true, then basically every website using this framework will get penalized, very less likely the case,
That's not how things work either, even different departments of Google won't add easily traceable watermarks that would get websites penalized,

People are mainly getting penalized due to mass publishing in general, which is most strictly moderated by AI,
Or due to the poor non-original content output, it's about the way you're using AI to get the output content in general,
And expecting to rank with low authority or/and new website(s), just because they "nailed" the content generation, which is just a part of the puzzle,
Lastly, UX signals, where most of the AI content will get you terrible stats, and they're being relied on heavily currently,

Idk for sure whenever Google is trying to directly penalize websites/pages using AI content,
But one thing for sure, poor non-original AI content made for the sake of "content length",
or/and to add content, with poor overall SEO (off/on) & spamming won't get you magically dominating whatever niche you're targetting,
You have to craft/find a unique way to use AI to get really good content that no one would tell it's AI, and invest into proper SEO, and that will surely work

 

[jyciap11]

Just a little bit of digging and here is what they're used for:
They're part of the "Tailwind CSS" CSS-Framework: https://tailwindcss.com/,
You can find similar codes in their docs here: https://tailwindcss.com/docs/line-height

If we're assuming your statement is true, then basically every website using this framework will get penalized, very less likely the case,
That's not how things work either, even different departments of Google won't add easily traceable watermarks that would get websites penalized,

People are mainly getting penalized due to mass publishing in general, which is most strictly moderated by AI,
Or due to the poor non-original content output, it's about the way you're using AI to get the output content in general,
And expecting to rank with low authority or/and new website(s), just because they "nailed" the content generation, which is just a part of the puzzle,
Lastly, UX signals, where most of the AI content will get you terrible stats, and they're being relied on heavily currently,

Idk for sure whenever Google is trying to directly penalize websites/pages using AI content,
But one thing for sure, poor non-original AI content made for the sake of "content length",
or/and to add content, with poor overall SEO (off/on) & spamming won't get you magically dominating whatever niche you're targetting,
You have to craft/find a unique way to use AI to get really good content that no one would tell it's AI, and invest into proper SEO, and that will surely work

Okay, I can agree with you in part on what you say, but I've made a thousand attempts (articles published en masse, articles published 1-2 times a day, 1-2 times a week, poor quality or high quality articles) and the only common denominator was these codes in the text.
The fact remains that “[]” is a piece of random code, meaningless, unjustified, and Google (I think) knows this.
Maybe I'm wrong, but this type of code is a particular feature of chatgpt, and Google knows it.

 

[hideath]

Okay, I can agree with you in part on what you say, but I've made a thousand attempts (articles published en masse, articles published 1-2 times a day, 1-2 times a week, poor quality or high quality articles) and the only common denominator was these codes in the text.

I made my fair share of tests too,
Without the same HTML attributes included, they're not included in any of the API content output, and plenty of test websites/pages got doomed,
Most of the websites & content (for parasites) that still rank are the ones made with hybrid approaches,
A mix of multiple specific prompts to refine it and with specific data/content provided to get well-crafted content,

Google could be trying to find and penalize whatever page/website using AI content, even while they're stating the opposite,
But that's not how it works, surely not using commonly used styling HTML attributes, that are not present in every content generated,
It's pretty much saying that Google will penalize every WP website just because a ton of spam websites have the same attributes in the source codes

 

[flatdeadsquirrel]

this has nothing to do with the html/css code. The reason content is being penalised is because it is being marked as Pure Ai because google uses Ai detection algorithms similar to the ones used by Originality [.] ai and loads of other Ai footprint detector tools out there. I.e Google, normally during major updates, refreshes the Ai score for each page and determines the likelihood that its 100% ai. I had an Ai site that was extremely well prompted discovered and bombed. It never had that code as we always copy and paste it vis notepad (i.e that strips any formatting)

 

[jyciap11]

this has nothing to do with the html/css code. The reason content is being penalised is because it is being marked as Pure Ai because google uses Ai detection algorithms similar to the ones used by Originality [.] ai and loads of other Ai footprint detector tools out there. I.e Google, normally during major updates, refreshes the Ai score for each page and determines the likelihood that its 100% ai. I had an Ai site that was extremely well prompted discovered and bombed. It never had that code as we always copy and paste it vis notepad (i.e that strips any formatting)

And simbol like this "[]" can't they be included in Google's AI detectors? Google bot analyze the code, of course, this is the first thing that do!

 

[MoonStarz]

ctrl + shift +v to paste without format as a plain text I used to have a lot of problems until I knew about it and you just answered a lot of people asking why their AI content never perform well they just lazy copy paste

 

[Proxy Cake]

Does anyone know an AI article cleaner to remove the watermarks?

 

[jyciap11]

Does anyone know an AI article cleaner to remove the watermarks?

Copy and past on Deepl. With Jasper works, remove all HTML/CSS useless but manteint correctly h1, h2 etc

 

[jyciap11]

I started new websites, same pattern (Jasper/chatgpt plus) but with the text formatted.
I will update the thread, zerpogpt give me 0-30% AI, so the content seems humans, I think I finally found why my websites got deindexed in a week or two!!
I will update the thread in the next month.

 

[Dream-z]

Thank you for the reminder

Directly copying HTML tags will expose that they are generated by AI

 

[MajestySEO]

I agree that blindly copying and pasting AI generated content Ctrl+C, Ctrl+V is not a good idea. AI tools are powerful, but if you do not edit, refine, or fact check the content, it can sound robotic, lack originality, and even get flagged as low quality.

Google does not penalize AI content itself, it penalizes low value, unhelpful content. AI can be a shortcut, but only if used wisely. The best approach is to tweak it, add personal insights, and make it actually useful for readers!