We recently launched a B2B website and published over 100 articles within the first month. Although our focus is not on the ranking of these articles, we intended to use the long-tail keywords in these articles to support the ranking of our service pages. Additionally, due to the silo structure approach, we planned to use these blogs as supporting pages.

Initially, the data looked promising, with many high CPC value keywords ranking well. I thought this strategy was working until it suddenly all went wrong within a month.

Boom! Everything went haywire!

The indexing of AI-generated articles has been continuously decreasing, with almost 80% of AI articles being dropped.

Updates: I created a new prompt to fight against that, you can read now or keep reading this article.

After this experience or lesson, I have stopped using AI detection tools altogether. It’s quite funny that there’s a comment on YouTube that says, “Use AI to trick AI and think that the content is not AI.” It’s like trying to deceive oneself that this is not an AI-generated article, which is similar to the Chinese saying “covering one’s ears while stealing the bell.”

Now, these AI-generated articles are mostly included in the “crawl but not indexed” category, making it difficult for them to be indexed. This is entirely different from being “discovered but not indexed.”

Some people have tried submitting URLs one by one, and some of them can be indexed, but they will eventually be dropped again after a short period.

What to Do If AI-generated Content Got Punished by Google?

To solve this problem, we should focus on several areas:

• Optimize the articles to improve their quality and readability.
• Use new URLs instead of the old ones.
• Remove the links to the old articles to avoid confusing the search engine’s judgment and wasting the crawling budget.

Therefore, our current focus is on emphasizing the context and readability of the articles. By doing so, we can improve the quality of the content and avoid any potential issues with AI-generated articles. This will help to ensure that our website is properly indexed and optimized for search engines, which will ultimately lead to better traffic and higher rankings.

Conclusion

Many people seem to be satisfied with the structure of the content generated by ChatGPT. However, when translated using DeepL, you can see that each sentence seems fine on its own, but when put together, the result is often stiff and unnatural. Even if it passes GTPzero.me, it may not pass Google’s BERT algorithm.

Google has been researching NLP for a long time, and trying to deceive Google with AI-generated content is like bringing coals to Newcastle.

Therefore, we now emphasize that AI is only an assistant, and we should focus on the coherence of the context. We usually rewrite the AI-generated articles to ensure their readability and coherence.

This is similar to doing SEO. No matter how fancy your technical skills are, if you don’t focus on solving problems for searchers, you won’t get good results. Similarly, Google emphasizes “helpful content,” not AI. Using AI to generate “helpful content” can still achieve good rankings.

For example, in an e-commerce project we are working on, our employees wrote an article titled “best…” that listed and explained more than ten categories of content, all of which were rewritten using ChatGPT. The article ranked very high because Google considered it helpful.

ChatGPT is a double-edged sword. When used properly, it can be an efficient tool, but if used improperly, it can be deadly.

I will continue to test our new strategy and share our progress if it is effective.