Tricks for LLM-generated pSEO

rocktheboat

Junior Member
Joined
Mar 18, 2025
Messages
115
Reaction score
48
Hey guys,

Wondering if any of you had tricks to share to make LLM generated SEO less detectable, and not be treated as spam

I'm writing my own scripts/agents for generating pages (which do research before writing) and already applying some post-processing like using a 2nd LLM step that remove obvious LLM words. But I'm sure Google is smarter than that.

Being also very careful to slowly publish new pages to not be treated as spam (currently 1 per day for my new domain).

Since this is consistently evolving field, very curious what's the latest best practices for that, and if you have any tricks that worked for you.
 
Have you tried to use third party SaaS and scripts for this?
I prefer to own the stack for that since it doesn't seem that complicated and integrates directly with my nextjs websites, which enables building cool stuff (like dynamic components in my articles with MDX)
 
Hey guys,

Wondering if any of you had tricks to share to make LLM generated SEO less detectable, and not be treated as spam

I'm writing my own scripts/agents for generating pages (which do research before writing) and already applying some post-processing like using a 2nd LLM step that remove obvious LLM words. But I'm sure Google is smarter than that.

Being also very careful to slowly publish new pages to not be treated as spam (currently 1 per day for my new domain).

Since this is consistently evolving field, very curious what's the latest best practices for that, and if you have any tricks that worked for you.
I have much more success when I limit the type of language that the LLM can produce. Limiting to a high-school reading level in the Flesch-Kincaid readability tests, limiting the maximum number of syllables that words should use etc.
 
Interesting thread. Are there any specific signals you've noticed that help reduce LLM footprint in the eyes of Google? Like structure variation, embedding multimedia, or even using hybrid paragraph swaps with human-written intros? Curious what’s actually moved the needle for others here.
 
Hey guys,

Wondering if any of you had tricks to share to make LLM generated SEO less detectable, and not be treated as spam

I'm writing my own scripts/agents for generating pages (which do research before writing) and already applying some post-processing like using a 2nd LLM step that remove obvious LLM words. But I'm sure Google is smarter than that.

Being also very careful to slowly publish new pages to not be treated as spam (currently 1 per day for my new domain).

Since this is consistently evolving field, very curious what's the latest best practices for that, and if you have any tricks that worked for you.
Make yourself a custom GPT and train it to write in a different sentence structure. Stock LLM output sentences are very wordy with lots of fluff. I've got my own GPT, then i throw it through a humanizer to get rid of the llm footprint even further.
 
For my short experience in SEO IA, the results I have achieved is to make the llms.txt and llms.xml files index in GSC, another clear factor is to analyze the niche I want to position and create backlinks in the sources that regularly uses chatgpt and gemini. perplexity is a good scraper to have the list of the most used and reliable sources; another data is to make the articles multi language.
 
Back
Top