Detecting AI-Generated Content: 6 Techniques to Distinguish Between AI vs. Human-Written Text
Trying to detect AI-generated content? Once you are familiar with what to search for, it’s not as challenging as it appears. Details inside.
In the ever-evolving landscape of…
Just kidding. But seriously. If you’ve seen an article or blog post starting with similar verbiage, odds are artificial intelligence (AI) is the true author of the text.
Undoubtedly, AI is disrupting nearly every industry in one way or another, driving the stock market to all-time highs.
Generative AI tools are helping companies, employees, and contractors streamline tedious processes. For example, freelance writers and bloggers who use AI say they spend 30% less time writing a blog post.
But this comes with a caveat. AI-generated content isn’t perfect. It often lacks style, personality, and emotion. And it can also get facts wrong and make things up, in a phenomenon now called Al hallucination.
While most would expect artificial intelligence to take the “word of the year” title in 2023, gee was the actual winner. Ironic, right?
If you do decide to use AI to write your content, it’s important to make sure it doesn’t feature the classic hallmarks of AI. Otherwise, you give away your content strategy within the first few paragraphs (or words).
So, knowing how to detect AI-generated content is the secret to striking the perfect balance between humans and machines.
6 Ways to Detect AI-Generated Content
Here are six simple ways you can detect AI-generated content from a mile away.
1. Proofread the Content
AI is highly efficient in producing content. But that comes at a cost. Often, the content is repetitive. There’s no real voice. And in some cases, the AI you use might make things up as it goes along.
That’s because generative AI tools like ChatGPT are trained on a huge amount of data dating back to a certain date. So, talking about current events or information after that date may produce inaccurate results.
That’s why it’s so important to proofread AI content. Seriously, don’t skip this step. All content should feature a natural, simple tone, avoid repetition, and provide accurate information.
Look out for obvious incorrect outputs, such as this attempt at an Amazon product description:
Instead of trying to save time using AI to write product descriptions, leverage WordLift to help you optimize your structured data to boost your chances of landing featured snippets.
Doing so will pay off in the long run and drive more traffic to your website. More traffic means more opportunities to convert leads into sales.
Now, let’s round out this first step. Your proofreading should go beyond just checking for grammatical mistakes. In fact, with AI, you’re likely to find zero typos and no grammatical errors at all.
Instead, you might find extremely fancy vocabulary with too much jargon. So, it’s important to look out for this, too.
That’s where proofreading services, primarily driven by skilled human editors, become invaluable. These services excel in identifying and fixing errors or inconsistencies that novice editors (or AI tools) might overlook.
2. Look for a Flat, Robotic Tone
Because AI writers are powered by, well, artificial intelligence, they lack a human voice. As a result, there may be a lack of personal opinion or emotion.
Let’s look at this example. Let’s say that you’re a digital marketing agency, and you ask ChatGPT to write two to three sentences on the importance of digital marketing.
Here’s the response it generates.
Screenshot provided by the author.
When reading this content, can you detect a personality or unique voice?
Probably not. The AI provides some pretty good information on digital marketing. By reading it, the average person can understand the true value of digital marketing.
But if you’re a brand that wants to get your message across in a way that makes you memorable and relatable, this is probably not content you want to share with your audience.
Why? It’s very monotone. Plus, it lacks emotion and depth.
Now, let’s look at an example of content with some personality. It’s the same topic but written in a more upbeat, relatable tone:
“We work, shop, and play in a digital world. You can’t afford to not use digital marketing strategies to get noticed and build brand recognition.
We’re talking strategies like social media promotion, search engine optimization, and email marketing to get your message across and let customers know your brand is here and here to stay.
And because you’re marketing your brand online, you can quickly adapt to changing customer preferences. How? Thanks to data-driven insights that help you continuously improve your strategies.”
See the difference? The brand is talking directly to the audience. It uses relatable language: “You can’t afford…”, “We’re talking,” and “Here to stay.”
While we’re on the topic of emotionless writing, let’s use HRIS software as another example. This software details payroll calculations and benefits packages, which may seem purely technical at a surface level.
Humans are the key to integrating that information with:
- Anecdotes about employee success stories
- The use of irony or everyday jargon
- Quotes from satisfied users
This personal touch, even in a technical context, goes beyond just conveying facts. It offers a human-centered picture of the software’s impact.
Why is that so important? Connecting with readers on an emotional level makes a lasting impression. And that’s exactly what an HRIS software company wants when emphasizing the human benefits of streamlining HR processes.
In short, a skilled writer can transform a dry manual into a relatable narrative, showing the value of the human touch even in AI-generated content.
3. Use an AI Content Detector
As we’ve touched on, your first task is to manually go through your content to make sure it doesn’t scream, “I was written by AI.”
Sometimes, it’s a little less obvious, and you need some help sniffing out that AI content.
Thankfully, you can also use an AI content detector to identify areas that feature AI-generated content characteristics, like repetitiveness and lack of tone and voice.
So, how do AI content detectors work?
They’re trained on human and AI-generated text to tell the difference between the two. But of course, they’re not always accurate.
Nonetheless, here are some of the characteristics they look for when detecting AI-generated content:
- Perplexity: This measures how predictable the content is. AI-generated content tends to have low perplexity. Human writing usually has higher perplexity, which results in more creative and complex language choices.
- Burstiness: This measures the variation in the length and structure of sentences. AI content usually has low burstiness, meaning there’s little variation in sentence structure and length. That’s because language models tend to predict the most likely word to come next, which makes the length of sentences and their structure more predictable, hence why AI can sometimes be monotone.
Of course, these traits aren’t always true for AI-generated content. Some AI writers are skilled at mimicking human language and tone.
This makes it difficult to detect AI-generated content, which leaves us in a gray area where we may easily mistake human-crafted articles for AI-made content and vice versa.
Take Cruise America, a Phoenix RV rental company, in their article “13 Travel Goals to Check Off in 2024.” Its crisp simplicity and practical information could lead one reader to assume a human touch, while another might suspect AI.
It can be tough to tell the difference. But AI-detection tools like Undetectable (Forbes’ #1 pick) can help you crack the code.
Screenshot provided by the author.
With a 90% accuracy score, according to Forbes, Cruise America passes the AI content test. The result? We can confidently say the text was written by a human.
4. Fact Check The Content
Distinguishing between AI-generated and human-written text is only getting more and more challenging as venture capital and investor money pour into this technology.
Now, advanced AI models can generate highly realistic and coherent content.
However, there are some more simple techniques to help make this distinction. We’ve already touched on basic proofreading. Now, it’s time to check for contextual understanding or unusual or inaccurate information.
For instance, we can apply some of these observations to this article on alternatives to Ozempic for effective weight management, which could be a candidate for AI-generated content due to the complex topic.
For context, here’s a screenshot of the article.
Here are some things to consider when trying to determine if the content is written by a human, using the above article as an example:
- Specific information and details: The article details Ozempic, how it works, who it’s for, how to take it, potential side effects, and its cost. This depth of information is typically associated with human-generated content.
- Use of citations: The article references percentages and information from clinical trials, suggesting a reliance on factual information. Proper citation is a common feature in human writing.
- Contextual understanding: The text demonstrates a reasonable understanding of the subject, discussing Ozempic and its use in treating Type 2 Diabetes and weight loss, referencing the current interest in the drug. This suggests a level of contextual awareness.
Whether you’ve used an AI writing tool and want to check your own work or you want to see if someone else has used AI, do a quick fact-check.
If you’re not an expert on that particular topic, you can leverage AI SEO Agent by WordLift. With its new ability to do fact-checking for you, you can validate claims and reduce the risk of incorporating incorrect hallucinations into your content.
This feature is game-changing because publishing inaccurate content can make you appear less trustworthy and alienate your audience.
5. Look for Repetitive Patterns in the Text
If you’ve used AI writing tools like ChatGPT before, then you’re probably familiar with how AI tends to repeat itself, but in different wordings or phrasings.
Screenshot provided by the author.
Notice how, in the above example, when asked to write a paragraph about eating healthy, the output from ChatGPT repeats the word “offer” or “offering” throughout the text.
Although the content is informative and shares some valuable tips, it repeats itself and doesn’t vary its word choices and sentence structure.
Remember, AI models are designed to be cautious and neutral in their outputs, which may result in more conservative language patterns. And this is what makes AI content sometimes look repetitive.
6. Run a Plagiarism Check
AI-generated content lacks the creativity and originality of human writing. That’s because it’s trained on content written by people all over the web.
As a result, AI writing may include identical or similar sentences from other publishers.
So, if you run a plagiarism check on a piece of content and it comes back with results, it’s possible that the content was AI-generated.
Screenshot provided by the author.
Learn to Detect AI-Generated Content to Build Brand Credibility
While AI content detectors are valuable tools, they aren’t 100% accurate. So, training the human eye to detect AI-generated content is crucial.
Key red flags are repetitiveness, lack of personality, and inaccurate information.
Sure, some AI-written text can pass as human writing. But you’ll become better at telling the difference when you know what to look for.
Use these tips and tricks we’ve shared today, and you’ll be able to detect AI-generated content from a mile away. Say goodbye to poorly written content and hello to engaging, human-written content that converts.
Happy editing!