Is AI Content Bad for SEO? (Answers & Facts)

If you want the short version of your question, the answer is yes. AI content does affect SEO. In fact, if done incorrectly, your entire domain could be penalized for the content that you publish and be removed from the SERPs (search engine results pages) for what we classify and call a “content penalty.” Finding out the reason why could be the gateaway to understanding how to rank anything well within Google.

Key Takeaways

• AI content on your website can have drastic negative effects on SEO. This is primarily driven by the fact that the score of information gain and the quality of your content is seen as “low” in Google’s systems.
• Modifying AI content does not increase its chances of succeeding for SEO. It will still lack key insights and uniqueness that are more heavily attributed to the information gain scoring system that Google has filed a patent on.
• Writing shorter articles, with more insights, can be a better way to spend less time on content writing instead of venturing into AI content, which can penalize an entire domain.

Google’s Experiments With LLMs and AI

The first thing you want to know is, why? Why does AI-content impact SEO? This might be your way of getting around it. If you can understand it, there’s a good chance that you might be able to avoid getting penalized or somehow rank.

Well, here’s why they know so much. Back in 2020, Google started to recognize the emergence of LLMs. They clearly started to put research time behind it. Enough so that they started publishing their own research papers.

Jason Wei, at Google, published a number of his findings here. In particular, many of these research studies started to show how Google was using the models to better train their own systems and improve them.

In the “Language Model Relevance Score” research study Wei writes, “Although automated metrics are commonly used to evaluate NLG systems, they often correlate poorly with human judgements. Newer metrics such as BERTScore have addressed many weaknesses in prior metrics such as BLEU and ROUGE, which rely on n-gram matching. These newer methods, however, are still limited in that they do not consider the generation context, so they cannot properly reward generated text that is correct but deviates from the given reference.”

Another Google engineer, Nan Du (who worked at Deepmind), has been playing with search-relevant applications for more than a few years as well. In the “Knowledge-aware attentive neural network for ranking question answer pairs” research paper it states, “Ranking question answer pairs has attracted increasing attention recently due to its broad applications such as information retrieval and question answering (QA). Significant progresses have been made by deep neural networks. However, background information and hidden relations beyond the context, which play crucial roles in human text comprehension, have received little attention in recent deep neural networks that achieve the state of the art in ranking QA pairs.”

The key takeaway? Google has been ahead of the LLM game for a long time. This isn’t something new to them. As a result, they utilized the newest technology to advance their systems. Now, put a pin in that thought for a moment.

Producing Content With AI and Google’s Information Gain Patent

If you think about what Google’s job is, it’s essentially to answer questions for you. You put in a question, you want an answer. The incredible task it has in front of it is to assume that this is the correct type of answer that you’re looking for.

A number of User signals can help to justify and evaluate their search algorithm. Ensuring that pages and queries are properly matching.

Patent #US20200349181A1 from Google. Classified as “Information Gain.” You can see from Google’s own model it has an “information gain scoring system” that it references upon the search of a query. It begins analyzing articles for presentation. And that’s when the information gain annotation engine takes over.

The goal of the patent and how the systems work? To recognize and surface new information that aligns with the latest “intelligence” that’s existing on the internet (or what’s being published). This is why Google can respond to news so much quicker and be incredibly targeted about assuming what you’re looking for when you search.

Is the storyline starting to make a little more sense? Let’s keep going and connect all the dots.

Why AI Content Doesn’t Work

It’s not that the content is AI written and that’s the reason why it doesn’t work. It’s that the insights that are contained within the writing are far too similar to everything else that’s already available on the internet.

For example, let’s say we did this:

• Produced 500 pieces of content with AI
• Produced 500 pieces of content by hand

Which is going to perform better? The reality: whichever one actually brings new insights to the table and addresses User questions in a more comprehensive way. In short, whichever one is actually higher quality content that’s uniquely taking a perspective on the subject matter.

So for example, if I had writers produce content that looked identical to another article, would I expect that to rank just because it was handwritten? Probably not. The same thing happens with AI-written content.

Want to work with us? Hire the best content writing agency for your tech business.

AI content detectors are not what you need

Tools like GPTZero and other platforms like this can help to identify AI-generated content, yes. However, “getting around the detectors” is not going to help you rank. You’re asking the wrong question if you’re asking this. That’s because Google’s systems have already long put in effort to identify what’s “regurgitated” and what’s “actually new.”

As a result of AI-boom, all they did with their helpful content update was simply calibrate the knobs of the ranking factors slightly more toward the direction of information gain rather than similarity (or relevance) of the content to the query.

Why AI Content Can Penalize Your Domain

In mathematics there’s a simple formula called a p-value. In short, “A p-value is a statistical measurement used to validate a hypothesis against observed data. A p-value measures the probability of obtaining the observed results, assuming that the null hypothesis is true. The lower the p-value, the greater the statistical significance of the observed difference.”

If you’re not catching on from the above, Google is scoring your page based on a number of factors, including:

1. How relevant your page is to the search query
2. How well you address the Users core question in the query
3. How much or many unique insights you bring into the topic
4. How well you display your expertise (usually a reference of entities)

As a result, they can give you a p-value of even just that simple page. However, in the mathematical formula, a p-value is a great way to compare other values against itself as well as create a cumulative total.

So for example, you could have p-values of a specific topic. Then you could p-values of domains that all speak about that specific topic.

Here’s why you can penalize your domain

Whether your content is handwritten or AI-generated, if it’s not making a difference in making the internet a truly better place, your p-value score (a mean average of scoring) of your domain will likely go down. As a result, you’ll be seen as “spammy” and potentially be removed from the SERPs entirely.

Poorly written content that’s poorly written by hand is the same as poorly written content that’s written by AI. You have to think about the intellectual depth of what you’re writing. And how well that’s empathizing with the core User that has questions about the subject matter you’re sharing.

Oh and did I mention that Google actually has a patent on this? Yes, they do, called the site quality score. Under patent #US9031929B1.

Summary

1. Google is not penalizing your content because AI wrote it. Google is penalizing your content because it lacks intellectual depth (AI simply remixes publicly available information, rather than "coming up" with new insights). This same method of analysis that Google performs against AI-content can happen with handwritten content (i.e., your handwritten content lacks intellectual depth).
2. Google can look at pages and determine whether the article(s) has new insights on an entity that it doesn't have within the system. It tests that content with low amounts of traffic to determine if the User signals are positive (clicks, engagement, bounce rates).
3. Google has been playing with AI (outside of Deepmind) for at least 4-years. Primarily, in the application of improving its search systems. As a result, it's not that they invented an "AI-detector," they simply invented a system that's really good at telling whether you're an expert in the subject you're writing about or not (experts tend to know more than the common person).

What to Do Instead

Most likely, you’re trying to save on the cost of content production. Our suggestion: start with smaller keywords that require less content. Or keywords where there is clearly a content gap. Thus, it is easier to rank.

You most likely don’t need more content, you most likely need a better SEO strategy.