AI's Ripple Effect on Google: The Evolving Landscape of Search Results

Learn what impact the explosive growth of AI-generated content has had on Search Results.

Be best prepared to navigate the sea of AI-generated content and understand how or if you should use AI content in your content marketing strategy.

The rapid rise of AI-generated content [1] has caused significant problems for search engines to index content.

It can also lead to a decaying feedback loop which we discuss in this post.

This has put in place an inflection point in how search engines deal with content and has necessitated a shift in search technology.

How has AI content affected search results?
Google Statements on Content
The Scale of the Problem
AI Content Detection
AI Content Case Studies
What About The Helpful Content Update
Theory on how Google treats AI content
What Does the Future Hold?
Summary
References and Resources

How has AI content affected search results?

Google has recently made many significant algorithm changes [7] that appear to be aimed, at least to some extent, at dealing with the significant challenge of AI content. From successive core algorithm updates, helpful content updates, and spam updates coming in short succession we have seen more SERP volatility [8] in the last few months than in the prior several years and that is saying something!

The rapid rise of AI content [2] and the webspam it has produced has led to this whirlwind of updates, leading to serious degradation in search quality [9] as Lily Ray points out in this LinkedIn Post.

Sistrix does a really good job of watching SERP trends, and volatility and finding the winners and losers of specific algorithm updates [23]. It is often the case that sites that get hit by some algorithm updates see a reversal at a later algorithm update.

Google Statements on Content

Google has made some conflicting statements but its general position on AI content [3] seems to be that it is perfectly OK to use AI-generated content.

Google changed its language on the use of human-generated content that seemed to give a free path to AI content (as long as it’s helpful). Read more on Google’s shifting approach to AI content. [2]

This seemed to give carte blanche to mass-produce AI content.

The Scale of the Problem

The problem is vast!

The web is vast.

Google’s original mission was to organize the world’s information. With an information revolution and now an AI revolution following, this has drastically changed the scope and scale of that mission.

Google can no longer ingest all of the content of the web.

It is too computationally expensive, even with Google’s size and scale. Google is investing heavily in infrastructure to the tune of $9.5B for data centers [11].

This is in part why new content does not get indexed. The other part is the helpful content update [10].

The bar has been raised for indexing, simply having a URL that returns HTML is not sufficient to get a page indexed.

Most of the guest post links you may pay for on pages will never be indexed and thus will never offer the value you may think you are getting for them.

AI Content Detection

AI Content detection [5] is entirely possible and changing just a few words does not work to hide the fact that text has been AI-generated.

The more AI-generated content you have on your site, the easier it is to detect [6] that it has been AI-generated.

There is a process referred to as AI watermarking [13] which involves defining a set of red and green words. This enables the text to be analyzed and can identify which set of words are more likely to have been used due to the probability of being aligned.

AI Content is done via LLM (Large Language Models) which use text embeddings [16] to represent text as a vector in a high dimensional space. AI text vectorization [17] is the process of representing text and turning it into text embeddings.

AI Content detection is a huge topic with various levels of testing and analysis. Some cases are better than others and some tools work better or worse in other areas. The article by Originality.ai comparing their tool to Content at Scale for AI content detection [20] is interesting, although you should be aware of the source and the scope of the analysis.

AI Content Case Studies

There is now a fairly significant body of case studies [18] that show that it is possible to generate lots of content that rank with AI. There is also a growing body of evidence that shows AI-generated content does not continue to rank [19].

One of the more recent and popular areas explored and analyzed was the SEO Heist [14]. This showed that it was possible to game Google and get a lot of rankings. There was much discussion about the SEO heist using AI [15]. This highlighted some of the issues in Google’s systems and has led to a rapid rise in web spam.

The economic argument for using a solution like Content at Scale [21] is clear with examples like this B2B SaaS use case [22].

SEO had been a cat-and-mouse game for a very long time and only in recent years has that shifted towards rewarding genuinely good publications, however the rise of AI content has certainly flipped the script.

Google is smart, Google will find a way to deal with this.

What About The Helpful Content Update

The helpful content update and its effect [4] on or treatment of AI content is interesting. Google has said it’s perfectly OK to produce AI content as long as it is helpful. If your content meets the bar of being helpful [10] and AI produces it this is seemingly OK.

My personal belief is that AI content should only be used to augment human-produced content and it should not form the main corpus of your site. Using AI to generate useful snippets and supporting content in certain sections of a site is OK in my opinion.

Theory on how Google treats AI content

My theory on how Google treats AI is that it has a post-process approach that looks at AI watermarking and penalizes or devalues content that is flagged as AI-generated, possibly impacting its site quality score as the primary metric that is modified.

It is too computationally expensive [12] for Google to do this at crawl time, hence the reason that some sites can gain massive traffic from AI-generated content. I believe there is a post-process mechanism that cycles through content with a specific algorithm and flags content that meets certain criteria.

We do see a rise and then fall with AI content and this can only be explained by some post-processing. If you consider that the index and algorithm changes don’t affect similar sites but sites with large amounts of AI content seem to have a life span of about 6 months before they crash and burn, this strongly suggests a factor/quality score adjustment/penalty by another name are at play.