ChatGPT got your tongue?

  • Updated: 02/12/2024
  • Published: 29/07/2024
  • 5 minutes read

ChatGPT got your tongue?

  • Updated: 02/12/2024
  • Published: 29/07/2024
  • 5 minutes read

With the introduction of large language models (LLMs) like ChatGPT and many similar tools, people have sped up their content production efforts. These tools found a place in several marketing divisions but really settled in affiliate marketing and content writing niches. Coincidentally, we have seen how Google Search results slowly devolve into low-quality yet heavily optimized spam.

A lot of AI-generated articles found their way to the top search results on Google. Articles that, although they sound great, actually have many factual errors or provide inaccurate information. Yet, they are heavily optimized and game the algorithm to become the most relevant. Affiliate marketing incentivizes the production of this kind of quick turnover content that serves one purpose – to get people to click and buy a product using a unique affiliate link.

Google responded to this surge in AI-generated content by issuing a core update to their algorithm, effectively reducing these websites’ ranking or even delisting them entirely. This over-reliance on AI-generated content has been flagged as spam. (Most delisted sites show AI-generated content to be at 50%, and some even climb to 90-100%.) 

What about reducing ranking? Some publishers reported a 60% drop in traffic from before the core updates.

Summary

Algorithm vs. algorithm

Google uses machine learning to identify which websites rely heavily on AI-generated content. The algorithm analyzes language patterns to identify whether the content has been created by a person or by an algorithm. If you’ve been using ChatGPT, you must have seen that every answer the tool provides is perfectly worded, which is not the case when people write. 

Beyond the language syntax, the search engine identifies and rewards high-quality content. In the past, the algorithm relied on keywords similar to how the index section works in books. But, people began stuffing keywords everywhere on their site – headers, footers, text and image descriptions. Some even went to reduce the text at the end of articles to font size 1 and color the text to be the same as the background. Nowadays, keyword stuffing can usually be found on poorly optimized e-commerce sites:

Keyword Stuffing Example. Source: Semrush

Then, links became a major ranking factor – again, similar to how references work in academia. For a time, website owners created high-value content and links were well-deserved. However, soon, people offered links to each other for a price or through link-for-link schemes. When Google figured it out, it penalized link trading.

Not long after, people realized they could create private blog networks (PBNs) and set up backlinks this way. Since then, they have incorporated PBNs into their SEO work, making it harder for people searching for information to understand whether the source is reliable.

Today, ChatGPT and other LLMs imitate natural language but often seem to miss the substance, get crucial details mixed up, or misrepresent facts. Because of these drawbacks, Google reacted to the emergence of AI-generated content by updating its core algorithm. Still, it didn’t stop people from using LLMs to replace writers. 

Relevance. Reliability. Access to information. These are some of the guiding principles Google has followed for its search engine. These principles have helped keep the internet somewhat reliable in the past, and they still hold when AI comes into general use.

Why this pushback for AI-generated content? From the perspective of someone who searches for information, people have become tired of looking for genuine information, not to spend their money. They don’t want to be sold to at every website they’re reading or every YouTube video they’re watching. Even adblockers are a thing, and have been since 2009. 

Now, with the surge of highly inaccurate texts that large language models like ChatGPT can create, internet users have renewed their need to find real-world experiences from other people, fully ignoring the AI-generated piles of text.

However, there is a twist.

Conclusion

One tool instead of the other - AI Overview

The twist happened shortly after Google updated its core search engine algorithm, giving precedence to high-value content created by people, not tools. Regarding the search engine, it was great to find that human-generated content has become more relevant than AI-generated content. Google noticed a surge in search prompts adding “Reddit” or “Quora” at the end. After all, when we search for information on the internet, especially about an experience, we want real people to tell us about it, not tools. We want to know more about a specific subject from someone who has spent time perfecting a craft or researching that subject. We simply want to find more relevant and reliable information.

Regardless, Google came up with its own AI tool – called AI Overview – that now analyzes the page and provides a recap of some of the results deemed acceptable by Google itself. As with all AI tools today, it misses the point, provides irrelevant content, or blatantly copies some answers without giving further value to the user. 

It appears that Google intentionally reduced the relevance of other AI content to boost their new AI Overview feature, which might replace Google’s featured snippets.

And as with any other AI tool, this one makes blunder after blunder. There have been reports that AI Overview gives incorrect, wrong, or dangerous information. Certainly, both Google Search and other AI tools will improve over time, and for now, human-generated content still reigns supreme. But will they replace people?