8 Best Ways How to Detect AI Generated Content

8 Best Ways How to Detect AI Generated Content

Until ChatGPT went mainstream, AI-generated content was mostly a novelty. While marketers and other niche users could leverage relatively crude algorithms for simple texts that defied the ordinary, the crafting of blogs and other long-form content was impractical. However, the release of ChatGPT and similar tools has made generating content with AI easier than ever.

Unfortunately for AI aficionados, the rise of AI has also ushered in a plethora of legal and practical considerations. For instance, the content might not be subject to copyright in many jurisdictions, and there’s more than one lawsuit over the use of copyrighted material to train algorithms. On a practical level, AI-generated content often lacks the quality, usefulness, and ranking potential of human content.

Because AI isn’t risk-free, businesses must be able to detect AI-generated content. Knowing the differences helps to protect your investment in quality content that drives sales and doesn’t make your lawyers worry about litigation. Plus, having full copyright for your content protects it from thieves. Ultimately, human-created content has a higher value, both practically and monetarily.

By following the tips in this article, you can learn how to detect AI-generated content.

1. Look for a Lack of Originality and Personality

YouTube video

AI-generated content often lacks a unique voice or perspective.  If there’s one thing humans still do better than machines, it’s thinking outside the box. We tend to analyze information and form independent ideas in ways that are, at least for now, out of the reach of AI.

But why do we think better than computers? Even the best AI models are trained on existing content, so output can sound generic. Essentially, these models produce new content from the building blocks they’re given and nothing else. Without additional training, those pieces of information will remain mostly static and less original over time.

Even with relatively generic information (like a quick how-to on eating sushi or getting a mortgage), human writers inject their personality, opinions, and style into their writing. Machines don’t crack original jokes or react to what’s happening around them. Similarly, conclusions reached by AI are logic-based and can’t benefit from intuition or natural inquiry.

Therefore, if the content seems bland, impersonal, or lacks a distinctive “flavor,” it may be AI-generated. The same goes for overly obvious or common conclusions.

Overly formal or inconsistent tone and style can be a red flag

The language style used can also indicate potential AI use. AI output may sound more robotic, stiff, and academic than a human’s casual, conversational writing style. In this case, you may also consider if the writer isn’t a native English speaker: often, people who haven’t mastered the language will also write with a formal tone.

Similarly, tone and writing style may also be inconsistent across the piece if generated by AI. Maybe one section is informal, while the other sounds academic. Or the text might use humor at one point while being dry elsewhere. Style swings are less common with non-native speakers, though, so this is an even bigger red flag.

2. Check for Factual Inaccuracies and Inconsistencies

Check for Factual Inaccuracies and Inconsistencies
Source

Do you think for yourself? Follow the evidence where it leads? How about using your intuition to filter fact from fiction?

If you said yes to any of these, you have an edge over generative AI.

AI can sometimes mix up or misstate facts

Ultimately, AI models are only as good as their training data. It is also hard for computer models to tell what set of facts is accurate.  AI models can “hallucinate” or fabricate convincing-sounding but inaccurate statements, as a lawyer learned recently after submitting a ChatGPT-drafted legal brief.

Factual errors are telltale signs that the content is AI-generated. While there are some misconceptions on the Internet – this used to be a reason teachers didn’t want kids to be dependent on websites for research – careful human review helps keep them out of higher quality content.

Remember, one reason we write blogs and articles is to gain a reputation as subject matter experts. Publishing erroneous content severely compromises this image. Always cross-reference important facts against reliable sources.

Further Reading: 9 AI Image Detection Tools for When You are in Doubt!

Look for inconsistencies in logic and reasoning

AI output can sometimes be self-contradictory or illogical. Young people who are still developing their writing skills may make these mistakes, but a professional knows how to keep everything consistent and well-organized. Humans excel at quality content with a narrative everyone wants to follow.

Logical leaps, non-sequiturs, or contradictory statements indicate potential AI generation. Most human writers won’t include these. In fact, I would consider the value of such poor content as nearly zero – both price-wise and in terms of the revenue it can bring in. Reading all content carefully will help you avoid paying too much for so little.

Further Reading: AI-Generated Content: Complete Guide to Its Pros, Cons and Best Practices

3. Notice Unnatural Repetition of Phrases

Have you ever learned a foreign language? I have, and it’s often a slow process. At first, you only know a few words, then simple phrases or sentences and canned responses. Luckily, people who speak the language better than you do understand you’re trying.

Is Your LinkedIn Not Delivering Results?

Just released: my new book to help professionals, entrepreneurs, and business owners maximize LinkedIn for real growth.

With years of LinkedIn expertise, Maximizing LinkedIn for Business Growth offers actionable steps to build your brand, expand your network, and drive results.

Start leveraging LinkedIn like never before—grab your copy now! Click the cover or button below to buy on Amazon.

Maximizing LinkedIn for Business Growth

AI models tend to repeat certain stock phrases and sentence structures in ways that are similar to language learners, but it’s often annoying. After all, your readers expect that the content will have a high standard of language.

How does AI show it isn’t a “native speaker?” The algorithm may overuse generic connective phrases like “in addition”, “furthermore”, “moreover, and other fillers. Unnatural and excessive repetition of words can also indicate AI use. Sentences may follow similar grammatical patterns and structures with little variation, just like a person who hasn’t learned a language well yet.

Humans use more varied vocabulary and sentence structures. Writers have a larger vocabulary and vary word choices more than AI. They’ll also mix up their sentence composition and cadence to create rhythm and flow, which avoids much of the “robotic” character of AI content.

Ultimately, more dynamic, less repetitive language indicates human authorship, as people naturally use a wider range of expressions and vary their linguistic structures.

4. Evaluate the Depth and Nuance of the Content

YouTube video

It’s one thing to explain a concept simply and clearly, but something else to be superficial. Unless you’re writing a children’s book, depth is important. Any human writer will understand this. On the other hand, AI-generated content may lack deep insights and nuanced understanding. It can over-simplify important concepts or indicate a lack of understanding. Similarly, AI struggles with nuanced language that indicates a writer knows what they’re talking about.

Analysis of content can also indicate probable human authorship. Look for evidence of critical thinking and expert knowledge that AI lacks. Content that reflects critical analysis, expert insights, or a strong personal viewpoint suggests human authorship.

Similarly, sections where the writer adds unique context or draws from personal experience are typically beyond current AI capabilities. If a piece of content has anecdotes and other creative features, you’re probably looking at a human writer.

Further Reading: Does Google Penalize AI Content? Here’s What We Know

5. Analyze the Flow and Coherence of Ideas

As individuals, we often judge content based on how well ideas flow and work together. Yet, AI can struggle with maintaining a coherent narrative. Look for abrupt topic shifts or a collection of loosely related points rather than a smoothly flowing argument. Quality content guides the reader through a logical progression of ideas, rather than jumping around to random facts or going on tangents.

Human-written content usually has intentional structure and pacing, even if the topic is informal, such as telling a story. A human author designs content to take the reader on a journey with intentional pauses, emphasis, and recapitulation. This way, readers get drawn into the content and feel like they MUST finish it. Otherwise, writers might “lose” their readers like the overzealous cousin who won’t stop talking about random topics at Thanksgiving dinner.

On the other hand, look for storytelling elements or rhetorical devices that enhance the reader’s understanding and engagement. Ideally, there should be something for everyone within the average human-authored content piece. Even adults can have fun laughing at the jokes in children’s books!

Further Reading: 15 Accurate AI Plagiarism Checkers to Ensure Original Writing

6. Scrutinize the Use of Sources and Citations

Scrutinize the Use of Sources and Citations
Source

While internet-based information sources are often much better quality than they used to be when students weren’t allowed to use them, this only applies to human-written content. Nowadays, we must be careful because AI may not reference sources appropriately or may invent non-existent references. Machines create content that “sounds” logical and knowledgeable, even if it’s anything but.

Before you post anything online, check the references and citations for accuracy and existence. Not only will you avoid paying for useless content, but being careful about sources safeguards your reputation. That’s because AI-generated content can mistakenly attribute facts or quotes to incorrect sources. Sometimes, the algorithm even invents facts.

On the other hand, human writing typically includes thoughtful source selection and correct attributions. People stake their reputation on accuracy and quality citations help ensure that the finished content is trustworthy.

Determining whether or not your content is up to human standards isn’t hard. Look for a well-curated list of sources that demonstrates the writer’s research process. Ideally, you should see sources from subject matter experts and reputable companies relevant to the niche. In addition, human writers often provide context about their sources, explaining their relevance or reliability. This is especially true with academic or professional niches.

7. Assess for Contextual and Cultural Awareness

As someone who worked in Asia for several years, I know how important cultural awareness is. For instance, using a knife at the dinner table would be frowned upon, while we Americans make knives an essential part of place settings. When people write content in a particular language or cultural context, they keep such conventions in mind.

AI may lack context or cultural sensitivity in its writing, which could result in awkward situations like making table knives part of the Japanese context. If content fails to recognize cultural nuances, idioms, or context-specific references correctly, there’s a good chance you have AI-generated content.

Likewise, AI-generated text might miss the mark on regional dialects or localized expressions. Languages like English and Chinese have several regional versions, sometimes even within the same country. In some cases, the same words can have different meanings to different localities. British and American English are notorious for this.

Human authors are more likely to display cultural competence and contextual understanding than AI. That’s because, much like Spock from Star Trek, the AI makes logical conclusions. Logic can be a blunt instrument that fails to understand nuance and conventions that fall outside its boundaries.

Ultimately, you should look for writing that engages with the cultural context in a meaningful way. This is much more likely to be written by a human.

Further Reading: AI in Content Marketing: Benefits, Ways to Use, and 5 Case Studies

8. Employ Detection Tools and Software

YouTube video

By now, you have plenty of tips on how to detect AI-generated content by analyzing the writing yourself. Nonetheless, with copyright and other legal issues still in flux, you may want to be sure that copy has plenty of human input.

Here’s one final check:

Utilize specialized software designed to detect AI-generated content.

As AI-written content proliferates, so do tools to detect it. Nowadays, we have a virtual arms race to see which technologies can fool the others. The developers of generative AI programs want their content to escape detection, while site owners use tools to minimize the chance that they’ll post a piece of AI-generated content.

Ironically, the very tools used to detect this kind of content are also AI-based. Options include Copyleaks AI, GPTZero, and Originality.AI. Each of these programs works a bit differently, so you’ll want to experiment with more than one to see what works best for you.

Understand that no tool is infallible, but the tools can be a useful part of a comprehensive detection strategy. In addition, you may want a human editor to comb through your content to detect errors before you publish it.

Finally, remember the limitations of current AI detection technology. Algorithms are only as good as their training data, and they might not keep up with the changes in generative technology. Although incremental changes are relatively easy to adapt to, leaps forward may give AI content providers an advantage.

Staying in touch with the AI game requires that you stay up to date on latest technological developments and detection methods.

Further Reading: AI Content Assistants: 13 Tools to Assist in Your Content Creation

Conclusion: Staying Ahead in the AI Detection Arms Race

The release of ChatGPT and other generative AI programs has turned the writing world upside down. While writing was a largely human task, this new technology has turned cheap content into more of a commodity. Even for human writers, there’s a temptation to use AI instead of traditional methods.

For this reason, it is more important than ever to think critically about a new piece of content. Gone are the days when we could just drop the text into Copyscape or a similar tool and then post whatever “passed” for not being plagiarized. You could even be sure the copy was eligible for copyright.

Using your best judgment and fact-checking skills will help you and your business only post quality, original work. AI doesn’t have the same level of expertise that humans do. They lack humor, cultural references, and other tools to make content relevant. On the other hand, human writers can access the full spectrum of human creativity, humor, and intelligence.

Further Reading: 15 Powerful AI Content Generators to Test Out

Fostering a proactive approach to content integrity

Ultimately, it’s the responsibility of every content reader to remain informed about AI-based content generation and detection. A combination of human and computer-based interventions will help you ensure that Web content is of higher quality.

Best of all, detecting AI content helps keep the Internet and its site owners relevant. We don’t want to go back to the Nineties and Aughts when teachers had to restrict students from including Internet sources in their writing.

At the same time, you’ll protect your brand and its collateral.

Hero Photo by Kenny Eliason on Unsplash

Actionable advice for your digital / content / influencer / social media marketing.
Join 13,000+ smart professionals who subscribe to my regular updates.
Share with your network!
Neal Schaffer
Neal Schaffer

Neal Schaffer is a leading authority on helping businesses through their digital transformation of sales and marketing through consulting, training, and helping enterprises large and small develop and execute on social media marketing strategy, influencer marketing, and social selling initiatives. President of the social media agency PDCA Social, Neal also teaches digital media to executives at Rutgers University, the Irish Management Institute (Ireland), and the University of Jyvaskyla (Finland). Fluent in Japanese and Mandarin Chinese, Neal is a popular keynote speaker and has been invited to speak about digital media on four continents in a dozen countries. He is also the author of 3 books on social media, including Maximize Your Social (Wiley), and in late 2019 will publish his 4th book, The Business of Influence (HarperCollins), on educating the market on the why and how every business should leverage the potential of influencer marketing. Neal resides in Irvine, California but also frequently travels to Japan.

Articles: 484
+
Table Of Contents