How to Detect AI-Generated Content
Table of Contents
The introduction of Generative Artificial Intelligence models is a truly significant technological milestone, as these models continue to revolutionize various industries in different ways.
Most commonly, these models help with content creation across different media, including text, music, and images, or employ their algorithms as AI website builders or grammar-correcting writing assistants.
This has led to a considerable surge in the amount of AI-generated content online, and the use of these technologies isn’t limited to content creation alone—even some newsrooms use AI. However, generative AI comes with some inherent flaws, such as bias, lack of understanding and context, reliability issues, dependence on training data, accountability issues, and misuse risks.
For example, just last year, video game fans fooled an AI-assisted website into writing about a fake in-game feature and presenting it as credible information. Other instances include AI producing socially biased content or even content that’s harmful to individuals and the masses. Therefore, using generative AI technologies in a responsible manner has become extremely important.
This brings us to the following question: How do we detect AI-generated content?
The aim of this blog is to explain how to detect AI-generated content, whether AI-generated content is detected and sanctioned by Google, and how to use AI tools to help with content creation without incurring penalties. So, let’s dive right in.
How to Detect AI-Generated Content
Generative AI relies on complex algorithms and machine learning models to generate various types of content, including images, videos, and, most commonly, text. So, when given a prompt, generative AI will use its training to predict and generate the text or visual elements that match the inquiry.
Of course, the specific methods and technologies used in the generative process vary based on the task and the AI model, but the working principle remains pretty much the same across the board. This means that the AI-generated content usually follows a particular pattern. We can use that pattern to discern whether a piece of content was created using AI technologies or not.
AI writing tools, such as ChatGPT, usually aggregate and process online information and paraphrase it to form a consistent and predictable output. However, unless given very specific instructions, the AI will, in most cases, provide you with generic information that’s otherwise easily accessible online.
Also, in some cases, the information might be repetitive and often contain irrelevant or even misleading facts. In short, AI-generated content like text is usually less complex than human-written text.
It may be oddly structured and formatted in either large bodies of text or broken down into fewer lines and shorter sentences. It also contains fancy vocabulary and very few, if any, grammatical errors.
Human writing, on the other hand, comprises varying writing styles, tones, and expressions and usually provides insights and facts that support the arguments. Humans love listening and telling stories, and human writers usually rely on stories and anecdotes to engage with their audience. On top of that, we’re more grammatically nuanced compared to pre-trained AI generative models.
The impact of image-generating AI has been ruffling several industries, especially since it has the potential to eventually replace live artists. However, as with AI-generated content like text, there are several key indicators that can help you recognize AI-generated images.
In most cases, images that are generated by AI have extremely rough or smooth textures, out-of-context background objects, surreal elements and landscapes, pixelated reflections, etc. This is particularly true when it comes to human features, which usually appear unnaturally smooth or even distorted—especially the eye gazes and fingers.
Additionally, most AI image-generating tools now add watermarks to aid easier detection. This helps companies and businesses discern what type of content their associates or employees deliver. Some companies actually encourage their employees to rely on AI for content generation.
The Importance of AI Detection
Considering the inherent flaws of generative AI models, businesses should be wary of using and employing AI-generated content in their online operations. To understand why this is, we have to discuss how Google and other search engines treat online content.
We’ve already dived too deep into Search Engine Optimization (SEO) elsewhere, but the short version is that Google relies on bots and crawlers to scan the contents of a particular webpage and rank them. That rank is subsequently used to rank that website on the Search Engine Results Page (SERP). The better a webpage ranks, the higher it’s positioned on SERP, leading to more conversions.
But if the content of the webpage offers generic and low-quality information with little to no value to those browsing or searching the web, Google will rank said webpage lower, favoring other, higher-quality content. Search engines love quality, domain authority, and thematic relevance, and if the webpage has all three, it will rank higher.
Using AI-generated content, which is usually lower in quality and value compared to human content backed up with insights and interesting facts, will rank lower. Since businesses want to make sure to publish relevant, optimized, and factual content, using AI-generated content could hurt the brand and waste company resources.
AI Detection Tools
Generative AI is becoming more complex and more sophisticated, and detecting AI-generated content “by hand” and “with the naked eye” is becoming increasingly difficult. That’s where AI detection tools come in.
These tools use a combination of machine learning and natural language processing techniques to scan a particular piece of content and try to determine whether the content was created by humans or generated by AI.
In short, AI-generated content detectors are AI-driven tools that try to detect and determine AI-generated content based on how predictable or unpredictable the content is and how varied or uniform it is.
But there’s a catch. AI content detectors, just like AI content generators, aren’t perfect tools. In fact, most tools will assign a likelihood based on their training to provide predictable results based on all the tested content. But these percentage values are mere predictions as well.
For example, if the results from a scanned text return 5% in favor of the AI, it doesn’t mean that 5% of the overall test sample was written by AI. It means that the AI detection tool would guess the sample was written by AI 5 times out of 100.
So, if the content consistently ranks under 10%, it’s almost certainly created by a human. However, once the percentage value gets close to 40% or 50% in favor of AI, business owners and managers should be suspicious about its origins.
Some of the most recognized and reliable AI-generated content detectors are:
GPTZero is an open-source detection tool that can be used to detect AI-generated content. It requires no signup; it’s free to use, though in a limited capacity.
Content At Scale is a cloud-based AI detection tool more oriented towards businesses, allowing them to detect AI-generated content in marketing materials and service interactions. It’s rather easy to use, and the free version provides the basic detection that’s quite adequate for most users.
Originality.AI is a top-tier tool that combines AI detection, plagiarism, and the use of paraphrasing tools, all combined in one cloud-based tool. Unfortunately for those seeking freebies, Originality.AI is an entirely paid tool ideal for marketing/SEO teams needing AI-free content.
Does Google Penalize AI-Generated Content?
Unfortunately, the answer to this particular question isn’t straightforward.
As explained above, Google relies on crawlers and bots to scan content on a particular web page, and SEO experts and content marketers have been trying to “game the system” for years in order to figure out what performs best on Google’s algorithms. In the end, better performance often equals more conversions, which equals bigger profits.
However, some SEO practices, such as keyword stuffing, diminish the value of the content, so Google set up a variety of both positive and negative factors to consider when ranking content. The company stated that it focuses more on the quality of the content rather than how the content is produced.
Google, as one of the companies at the forefront of AI development, actually encourages the use of AI and AI-generated content. However, while the company has clearly stated that the appropriate use of AI or automation isn’t against its guidelines, using such content to manipulate search ranking is against Google’s spam policies.
The bottom line is that Google rewards high-quality content regardless of how it’s produced and penalizes poor-quality content, regardless of whether it was produced by humans or generated by an AI.
Generative AI Isn’t a Bad Thing
Just like any other technology, generative AI isn’t a bad thing.
It’s actually pretty useful, and many industries have been using generative AI to do the heavy lifting. For example, the gaming industry relies on AI for asset creation, content marketers rely on AI for content ideas, and web design companies rely on AI website builders to do most of the heavy lifting.
The latter is particularly interesting for several reasons. Owning a website has become a necessity in modern business, and it’s particularly important for small businesses as it increases their visibility online. However, not all businesses are backed up with massive budgets that would allow companies to contract web designers.
So, many business owners rely on AI website builder tools to build their own websites and establish their online presence.
Using AI website builders eliminates the need for time-consuming coding, overall development, uploading to hosting, and endless troubleshooting. Not to mention that these tools provide a rather cost-effective solution without compromising quality.
The possibilities are near-limitless because most AI website builders, such as Hocoos, come equipped with various options, such as marketing integrations, time-saving email campaigns, blog platforms, online stores, and a vast majority of other options appealing to small service business owners.
And since Google doesn’t penalize AI-generated content, there’s no reason for websites created using AI website builders to be negatively affected. That is, as long as they’re adequately structured and subsequently optimized for search engine indexing.
AI website builders aren’t the only example or most commonly used AI-driven tools. Grammarly, a well-known and popular digital writing assistance tool, also relies on artificial intelligence and natural language processing to help users with grammar, punctuation, spelling, word choice, etc. In fact, the vast majority of professional writers use Grammarly for proofreading and grammar checks.
With that said, it’s important to note that the use of Grammarly in text might increase the AI-detection score by a negligible margin since Grammarly relies on GPT models.
How to Use Generative AI Without Incurring Penalties
We already discussed that Google doesn’t actually penalize AI-generated content of any sort for as long as that content provides valuable and insightful information. Considering that most generative AI models only produce basic, robotic content with recognizable patterns, the use of such models for generating the entirety of the content isn’t advisable.
For example, having ChatGPT write an entire blog post or article will likely result in flagged content, which can adversely affect the company website’s SEO and SERP rankings. However, that doesn’t mean that AI has no use in the creative process.
When it comes to creating content, the key is to create high-quality content for your target audience instead of simply “gaming the system” to boost your ranking to page one of Google Search. When it comes to written content, the process is a bit more involved than writing alone.
One has to conduct SEO research, find credible sources of information and citations, and then write the content. AI can help with the workload, provide certain information on topics the writer lacks a clear understanding of, and even produce the content.
But for the content to be deemed high-quality by Google, it has to be adequately formatted, its claims have to be backed up with credible data—or even better, studies and surveys—and then edited for grammar, tone, and voice.
Generative AI can help speed up the workflow, often structure and transition ideas, and ensure that the grammar is on point. However, it sometimes misunderstands context and can provide inaccurate corrections, which is why it should be reviewed by a human editor.
In the end, creating high-quality content comes at a price, especially if we factor in the expertise required and hours spent, regardless of the project, niche, or industry. This also applies to image-generating AI, AI writing tools, and even AI web builders.
The difference in our approach to generative AI and its use can make a difference between a successful business and a complete waste of a company’s time and resources.
AI-generated content can be damaging to a business or a company, as it often provides basic or inaccurate information, further underlying the necessity for better AI detection models.
However, embracing AI technologies as a tool of modern business provides businesses with the opportunity to leverage the power of AI for content generation and enable professionals to produce better, high-quality content.
How to Detect AI-Generated Content
Detect AI-Generated Content with GPTZERO
(1) Head to GPTZero
(2) Pase your content into the text box provided (or upload a doc)
(3) Click “Check Origin”
Increase AI Detection Quality and use accounts by signing up to a paid plan, the cheapest of which is $10 a Month
Detect AI-Generated Content with Content At Scale
(1) Head to Content At Scale
(2) Paste your text into the provided text box
(3) Click the button that says “Check for AI Content.”
There’s a limit of 2,500 characters, so if your text is longer, you’ll need to check it in sections. If you want to perform checks on bigger bodies of text, you can sign up for the premium version.
Detect AI-Generated Content with Originality.Ai
(1) Navigate to Originality.ai AI content checker page (also accessible from the homepage navigation bar)
(2) Paste your text into the designated field on the page.
(3) Click the “Check Originality” button to analyze your text.
Character and usage limits are low, with a basic account starting at $15 per month.