Artificial Intelligence is changing everything in the business landscape. Content creation is also not untouched by this force, and AI-writing tools have emerged now that can generate content looking as good as the content written by humans.
However, forming sentences that look like they were written by humans and putting them together to create a blog post is one thing; creating posts filled with elements of human experience and expertise is another. That still can’t be done by AI writing tools. Google also realizes that, and that’s why it has made clear that AI-generated content won’t be tolerated in search results.
In such a situation, it becomes critical for you to have the necessary tools and skills to determine if any of your writers are using AI to generate your content. Because if they are, then you may have to pay for it in the form of a search engine penalty in the coming days. That’s why in this post, we’ll explore how to detect AI-written content. Let’s begin!
There are 3 methods to detect AI-generated content: Manual scanning, the use of AI detection tools, and the use of Chrome extensions that use AI detection tools.
Manual scanning relies on looking for certain patterns in the content. AI detection tools, on the other hand, have their own algorithms, but their reliability is still a matter of debate. Four popular AI writing detection tools and their reliability is as follows:
|Tool||AI Content Detection Accuracy||Human Content Detection Accuracy||Reliability|
|Writer.com AI Detection Tool||1/3||4/10||Less reliable|
|Copyleaks AI||3/3||8/10||Quite reliable|
|OpenAI Text Classifier||3/10||9/10||Partially reliable|
Chrome extensions of these tools, on the other hand, can be used for situations when you want to check a lot of content that has already been published.
#1. Pay Attention To Patterns
A lot of times AI-generated content can also be detected by taking a mere look at the content in front of your eyes. When you scan through it and read it, if you observe patterns between sentences, then there’s a good chance that this content was generated using AI. Here are some common patterns found in AI-generated content:
- Repeated use of certain terms/phrases
- Identical beginnings of paragraphs/subheadings
- Identical introduction/conclusion format in multiple posts
- Excessive use of complex synonyms instead of more common counterparts
If you notice more than one of these things in content supplied by any of your writers, there’s a good chance that they may be using AI to generate your content.
#2. Use AI Detection Tools
There are a number of AI detection tools that can help you identify whether or not an article, blog post, or any other form of content was generated using AI.
The first tool on our list is Originality.ai, which includes both plagiarism detection functionality and AI detection. At $0.01 per 100 words, it’s a bit pricey in comparison to other plagiarism checkers, but ultimately it’s the only tool that you’ll need for your content quality needs after you purchase it.
As far as accuracy is concerned, the Originality.ai team says that their tool can detect content generated by GPT-3 and GPT-3.5 based tools (i.e. ChatGPT and other online content generation tools) with up to 94% accuracy. However, this accuracy requires you to input at least 50 AI-generated words in order to detect the use of AI, which means it may not be able to detect AI footprints in certain types of content, which include:
- Ad copy for Facebook and Google ads
- Meta descriptions
- Product descriptions
- Some introductions and conclusions of blog posts
All these forms of content are often shorter than 50 words, so if someone used AI to generate them, then Originality.ai may fail to detect the use of AI in them. Regardless, it’s still a pretty useful tool, especially for blog posts and articles.
Originality.ai is also the most reliable AI detection tool on our list, as it passed with a 100% success rate when tested with 3 prompts of AI-generated content by Search Engine Journal and with a 90% success rate when tested with 10 prompts by Artturi Jalli of BloggersGoTo. However, when tested with human-generated content, the results were not quite reliable as it could identify it as human-generated in only 4 out of 10 cases.
2. Writer.Com AI Content Checker
Writer.com is an AI-based content generation tool itself, but they also provide an AI content checker that can help you detect AI-generated content. You can search using both a URL as well as text copied and pasted into a text box, and upon performing a search, the tool will let you know if the content has AI footprints or not. The best part is that it’s completely free to use.
However, the free cost comes at a price. Individual reviewers like Artturi Jalli have found in their testing that 6 out of 10 times, it can flag even human-written content as AI-generated. And Search Engine Journal discovered in their testing that 2 out of 3 times, it failed to detect AI in ChatGPT-generated content even! We, too, had a similar experience in our testing.
In short, it’s free to use but not a very reliable choice to detect AI content. You should combine it with other methods (i.e. manual scanning, which we explained in the first step) to arrive at a reasonable conclusion regarding the use of AI in any content piece.
3. Copyleaks AI Content Detector
Just like Originality.ai, Copyleaks is actually a plagiarism detection tool, but it can detect content generated too. It claims that it can detect content generated by a wide range of AI-based content generation tools, including ChatGPT, GPT-3, and even content that has a mix of human and AI-generated content. However, it doesn’t allow you to check the content of a webpage using a URL. You have to paste the content in it, only then it can do the job!
As far as accuracy is concerned, SearchEngineJournal found in its testing that it’s pretty reliable despite being a free tool. In all 3 prompts when it was supplied ChatGPT generated content, it detected the use of AI. However, they didn’t perform a testing of human-generated content on it. When Artturi Jalli performed those tests, he found that it made only 2 mistakes out of 10 in flagging human-generated content as AI-generated.
So it seems to be a reliable tool, and a pretty impressive one actually considering the fact that it’s free, but still, a 20% false positive rate for human-generated content means that we should take it with a pinch of salt!
4. OpenAI Text Classifier
Content generated by ChatGPT is a big challenge for all publishers seeking genuine content, and ChatGPT developers also realize that. That’s why they made available this tool called OpenAI Text Classifier, which can help you detect ChatGPT-generated content that’s often hard to detect for other AI detection tools.
However, it’s worth noting that this tool is only partially reliable. It has been pointed out by the OpenAI team itself that in their tests, it generated 26% of true positives and 9% of false positives. In simple language, out of all AI-generated content, it could detect AI footprints in only 26%, while it also flagged human-written content as AI-generated in 9% of cases. So use it with a healthy dose of skepticism.
#3. Use Chrome Extensions
There also are a number of Chrome extensions that can make detecting AI-generated content quite easy, especially when you are dealing with lots of content on multiple web pages. Some popular ones include:
These extensions can be triggered simply by clicking a toggle button from your Chrome address bar. Once triggered, they can scan the content on the web page to tell you whether or not it was generated using AI.
Conclusion: How To Identify AI Written Content (2023)
Generated content and AI detection tools both are evolving technologies. AI-based content generation tools are evolving every day to acquire the capabilities of generating content that’s on par with human-written content, while AI detection tools are racing against them to get smart enough to detect such content.
In short, this whole space is developing and growing at a breakneck pace, and when that happens with technology we should use that technology with abundant caution. If you’re a writer, you can’t rely solely on AI-based tools to generate all your content for you. Similarly, if you’re a publisher, then you can’t rely on just one or two tools to detect AI-generated content with 100% accuracy. You’ll have to apply your own common sense as well.
What it means is that if you want to detect AI-generated content, use your own expertise and editing skills in addition to AI detection tools to arrive at any conclusion. A human editor and experienced writer ar still quite valuable job roles (in fact, more valuable today than they have ever been), and AI is still not in a position to replace them.