As AI writers (like SEO.ai) and content generators become more and more widespread, so does AI-generated content.
This leads to the natural question; Can Google can detect AI content?
The answer is ambiguous. Yes, they might be able do detect some AI content. But not all. And Google might not be able to do this in the future as tools like paragraph rewriters are readily available.
And would they want to use a detection algorithm as part of their rankings and can you then avoid detection?
Let's dive.
What is AI content?
Let's start with the beginning and look at what AI content is.
AI content is content that is created using Artificial Intelligence (AI) technology.
This type of content can range from simple text, to complex visuals, to interactive experiences.
AI content is created by algorithms that are programmed to generate content based on user input or data.
This is also often referred to as generative AI.
Several types of AI content can be generated by generative AI.
- Text: such as articles, stories, poems, product descriptions, and chatbot responses.
- Images: such as photographs, illustrations, and infographics.
- Audio: such as music, speech, and sound effects.
- Video: such as animated videos, commercials, and film scenes.
- Voice: such as speech for virtual assistants, narration, and voiceovers.
- Natural Language Processing: such as language translation, sentiment analysis, and summarization.
- Virtual Reality and Augmented Reality: such as 3D models, environments, and animations.
- Code and Programming: such as software, apps, and games.
- Financial Data and Predictions: such as stock predictions, financial reports, and risk analysis.
- Scientific Data and Predictions: such as weather forecasting, scientific research, and drug discovery.
In most cases, when SEOs discuss Google and detection, they are thinking of whether a texts can be recognised as AI generated.
But images, videos etc can also be generated by AI (the illustration for this article was made by Midjourney) and is part of the ranking factors of standard web content.
The same video and we see AI video of all sorts that is gaining more and more progress.
And this can also be used for SEO purposes in a more automated or spammy way.
An example; Someone used a combination of AI-generated text and images to drive traffic to their website.
They created tens of thousands of images with various variations using tools like interiorai.com and posted them on Pinterest.
Each image then had a link to the creators site in the image description.
How can Google detect AI content
If we are focusing on the text made by AI's there are different ways that Google could try spot a text coming from an AI system.
To understand how, it's first essential to understand how AI can generate text.
AI-generated text is created using a process called Natural Language Generation (NLG).
NLG is a type of Artificial Intelligence (AI) that uses algorithms to generate human-like text from data.
The algorithms are trained on a large corpus of data and then generate text that is similar to the data.
Google can also use machine learning algorithms to detect AI-generated text by looking at signals such as the text structure, grammar, and syntax in the text.
In other words these algorithms can detect patterns in the text that are indicative of AI content.
Example text:
"The new iPhone is set to revolutionize the smartphone market with its advanced camera technology and sleek design. The device features a dual-lens system and advanced image processing algorithms, making it the perfect choice for photography enthusiasts. With a spacious battery and lightning-fast performance, the iPhone is sure to impress."
This text could be identified as AI-generated by Google in several ways:
- The structure of the text is repetitive, with similar phrases and words used multiple times.
- The grammar and syntax are correct, but may lack the natural variation and complexity found in human-written text.
- The text may have a high degree of coherence, but lack the nuance and context found in human-written text.
What we also see many low-quality AI writing tools just produce large walls of text.
Or if someone is not combining AI with human interaction when writing.
This is just bad content, and it being downgraded by Google have nothing to do with it being generated by AI, but because it's simply low quality content that is not having a good user experience.
Why would Google try to detect AI content?
What would be the motivation for Google, as a search engine, to detect whether the content is generated by AI or a human?
Google is striving to give users the most optimal search results and is constantly working to enhance this process.
A major component of this optimization is to stop displaying spammy content.
And up until now, a major contribution to spammy content has been "automatically generated content".
The cause of this is that the output from various types of scripts has been of very poor quality, consisting mostly of keywords and being used to try to manipulate Google's search engine rankings.
Google has been attempting to identify this kind of content and take it out of the search results (SERPs) in order to maintain the accuracy of its search results.
In Google's guidelines, they define it in this way;
Spammy automatically generated (or "auto-generated") content is content that's been generated programmatically without producing anything original or adding sufficient value; instead, it's been generated for the primary purpose of manipulating search rankings and not helping users.
What is important here is the distinction between "Spammy automatically generated" and "AI content".
With the improvement in technology, where AI output can be as good (or sometimes better, I argue) it's does no longer make sense to be against things generated with AI per default.
Also, there are lot of grey zones.
Eg. this text you are reading now have been made in a combination of AI and me as a human.
Does that make it AI content?
Or what about something written and then improved grammar with an AI tool like Grammerly?
Hence the change in the Google Search Essentials (former Google Webmaster Guidelines).
Now Googles focus is on whether it's spammy and in general helpful content.
Why would you want to avoid Google detection?
We still see a lot of discussion on how to avoid Google detecting AI content.
But what would be the reason for trying to mask that a machine has been involved in the creation?
I believe it can fall into three different categories;
- SEOs that still believe Google is against AI content (which is not the case anymore)
- SEOs that are skeptic and feel safer if Google cannot detect it as AI content
- SEOs that use AI for spammy purposes
By trying to conceal that generative AI did output the text, SEOs falling into one of the 3 categories feel safer that they are not getting a penalty by Googles ranking algorithms.
What I hear from time to time is also that some, mainly in group 2 above, fear that even when Googles does not penalize today, they might do this sometime in the future.
But as described in the previous section, it makes less and less meaning to downgrade content depending on whether an AI has been involved in its generation.
The AI will just keep getting better and better (meaning improved quality and helpfulness) and the lines between AI/humans will be more and more blurred.
If something the trend is going in the other direction. Just like we have seen Google go with it's update in guidance/stance on AI.
4 ways to avoid detection
So if you are falling in into one of the three groups, how could you then mask your content to be sure that Google cannot detect it?
The three different ways to do this we see people try;
- Get the AI to output text less likely to detect
- Use another tool to camouflage the text as AI generated
- Use a tool with a more advanced AI engine
- Mixing human and AI writing
You can also look at our small experimentation with a text in 10 different AI detector tools.
1. Get the AI to output text less likely to detect
The initial way to mask the text, can be to give different instructions to the AI that will output text that is looking less like machine generated.
This ability is especially something the newest language models are good at.
By using more advanced language generation techniques, such as those that take into account context, tone, and style, the generated text will be less predictable and more difficult for other machine learning algorithms to detect.
One can also use prompts that provide more emotional context:
Instead of providing prompts that lack emotional context, provide prompts that provide more emotional context, such as "Write a letter from a mother to her son who is leaving for college"
Or use prompts that provide more creative context, so such as "Write a poem that describes the feeling of falling in love".
In my small experiment with different prompts and detector tools, I got the best result with the prompt "Rewrite following as a 24 year old human would;"
2. Use another tool to camouflage the text as AI generated
Another popular way to camouflage that an AI made the text, is by having the text rephrased by another AI writer.
Often its the Quillbot that is referred to as the go-to platform for this.
This strategy seems rather odd to me.
Taking the text from one AI and having another AI rewrite it.
This should only be an improvement if the initial tool or initial output by the first AI platform is of low quality. And the Quillbot is then instructed to rephrase in a specific manner. Like my example of rewriting as a human at a certain age.
3. Use a tool with a more advanced AI engine
As the language models evolve, it will output better and better text quality.
This also makes it more difficult to detect, as there will be greater variance in the text and it will seem more human.
So the more you use tools that use more advanced models such as GPT-3 (both SEO.ai and ChatGPT uses GPT-3.5), which can generate more human-like text, the less likely it is that it can be detectable by other machine learning algorithms.
And when the GPT-4 is being introduced sometime this year, it will just increase.
4. Mixing human and AI writing
The last way to avoid detection is to mix human and AI writing.
This means that you can take a piece of text that has been written by an AI and have a human writer edit it to make it more natural.
Or you can do it the other way, and utilize AI as you write your article to speed up the process, but still, add your own parts.
This is the way I prefer.
This way, you can get the best of both worlds: the speed and accuracy of AI writing, and the naturalness of human writing.
By mixing human and AI writing, you can make sure that the content looks more natural and less machine-generated.
This is a great way to make sure that your content is not detected as AI generated.
And it's easier to be aligned with the latest "Experience" factor introduced by Google in its E-E-A-T (Experience, Expertise, Authority, and Trustworthiness) for rating content.
Key takeaways for SEO professionals
So what is the conclusion;
Can Google detect AI content? Yes, most likely they can detect some content is being generated by AI.
Will they be able to make AI detection in the future? Most likely not, as the AIs evolves and become better and better and imitate humans more and more.
Should you be worried as an SEO using an AI writer? Not if you are focusing on generating helpful content with the users in mind.
And we recommend you do not just auto-generate all text (it can be perceived as spammy) but mix manual human writing and AI.
.svg)
