The arrival of OpenAI’s ChatGPT has caused a stir in the world of AI. While many applaud the advancements in natural language processing, others are concerned about how much authenticity will be left in the written word when AI-generated articles become commonplace.
AI-generated articles are indistinguishable from human writing in many cases and can be used to spread false information or manipulate public opinion. To protect against such potential misuse, people are discovering ways to detect AI-written content.
We’ve gathered the most helpful tips around the web and boiled them down into a simple, step-by-step guide. Here’s how to detect AI-generated articles:
With the quality of results produced by the latest language models, it can actually be quite challenging to detect AI-generated articles. However, with a keen eye and some careful inspection, it’s possible to spot the telltale signs of automated writing.
Perhaps the most obvious sign of AI-generated content is repeated words, phrases, or sentences. AI is designed to recognize patterns and replicate them as accurately as possible. As such, you may see the same sentence structure used regularly in different paragraphs within the same piece of writing.
AI-generated articles may appear coherent at first glance, but if you look closely, there are often structural issues. For example, AI-written content tends to be overly formulaic and lacks natural transitions between topics or ideas. They are also not mindful of grammar rules or basic punctuation.
Depending on the topic, an entirely AI-generated article may only be written in the general sense. What we mean is that it could gloss over facts and figures. So when you read an article, notice the details (or lack thereof).
AI language models are not continuously updated. The latest OpenAI models are only trained on data from 2021 and back. So if the article you’re checking deals with recent facts and figures, it can be difficult to detect hints of AI.
Still, AI writing tools are designed to generate believable facts and statistics. If you notice any inconsistencies in what’s being reported, it could signify that the article was produced by an AI.
Another telltale sign of AI-generated content is a lack of emotion or personality in the writing. AI algorithms are built to be as efficient as possible and replicate human writing with minimal effort, so they may omit emotional words and phrases that are integral to natural language expression.
However, the latest AI language models are better at mimicking human writing, so this tip may not be as reliable. Be sure to look for other signs as well.
If you can’t easily determine whether an article is AI-written or not, look into its author’s background for more information. A quick Google search could reveal any past work they’ve done that would suggest they are capable of producing the content in question.
No traceable background? The author’s past work does not match up with what you’re reading. Then there’s a possibility it could be written by AI. But this shouldn’t be the only factor to consider since many experts hire ghostwriters for their work. It’s important to do your due diligence and check for other signs.
The manual inspection methods discussed above will help you detect AI-generated articles to some degree, but they are far from foolproof. For more accuracy, you’ll need to rely on specialized tools and services.
Originality.ai is trained on OpenAI’s models and is known to be highly accurate in its AI detection capabilities. According to tests, the tool can detect AI instances in content produced by GPT-3, GPT-3.5, and ChatGPT. It also takes a two-pronged approach to detecting AI-generated content.
First, it determines whether or not a text is AI-generated. Second, it checks the content against millions of web pages for plagiarism. This ensures that the content is both original and authentic.
Here are step-by-step instructions on how to detect AI-generated articles with Originality AI:
Unlike other AI detection tools, Originality.ai requires users to sign up for an account. Originality.ai is not free to use, but it does offer a free trial through its Chrome extension. If you want to try it out first, simply go to the Chrome Web Store to download it.
On Originality.ai’s dashboard, you’ll see the tab “Content Scan.” Navigate to this tab to paste your article. You’ll see how many credits the article costs. If you’re okay with the number, click the “Scan Now” button.
Originality AI’s results display two components. On the left is the AI detection score, shown in percentages. It will show the percentage of the article that is original or human-written and the percentage of it that is AI-generated. While Originality.ai is great with accuracy, it is not 100% accurate.
The second component is the plagiarism score. This is also quite helpful because although articles can be %100 human-written, it doesn’t necessarily mean they’re plagiarism-free.
In the era of AI-generated content, it’s important to be vigilant about what we’re reading. While it can be difficult to detect AI-written articles, with a little bit of effort and attention to detail, you can usually spot them.
To be more sure, you can use online tools like Originality.ai to verify the authenticity of articles and other written content. These detection apps are trained on AI-generated content, so they can provide a more reliable assessment.
However, AI content detection tools are not to be used as the sole substitute for critical thinking. The best way to detect AI-generated content is still by using your own judgment and paying attention to the details.
