From robot to real: 4 ways to detect AI-written content in seconds

As if (legit) content writers and content marketers couldn't roll their eyes enough at the people who thought writing good marketing content was as simple as throwing a grammatically-correct sentence together and clicking 'Publish'... in came generative AI; the easy, breezy, unbelievably-accessible tool that anyone can use to create content.

Next thing we know:

• Every post on LinkedIn is talking about some kind of "fast-paced landscape"
• Companies are bombarding the internet with 300x more (disconnected) blogs a second
• Readers are dangerously digesting content that isn't necessarily 100% factually-correct

The fun, engaging, and authentic personalities of human-written content are fading into a sea of AI-generated content... faster than Jack off that floating door in Titanic (and maybe even evoking more emotion to us content marketers, too).

But all is not lost—unlike Jack in the North Atlantic Ocean. There are ways to detect AI-generated content; whether you're trying to cover your tracks, protecting your content strategy, calling someone out, or... just plain bitter. 👀

Why you don't want purely AI-generated content

Sure, speed is a huge benefit of using AI for content creation but when almost everyone is doing it, the cracks start to show.

By going AI-only with your content, you can see problems with:

• Accuracy and quality
• Tone of voice or brand voice
• Lack of human emotion
• Zero-to-low creativity
• No new insights or originality
• Search rankings

☝️ All of which do absolutely nothing for audience engagement.

The cure? Human oversight and intervention. More on that in just a scroll—but first, here's how to spot AI-generated content in seconds.

4 ways to detect AI-generated content in seconds

1. Linguistics patterns and markers

   
   Okay, so you may already have realised that "in this fast-paced landscape" sends a shiver down my spine, but there are a bunch of overused phrases and AI clichés that you are guaranteed to see if you copy-paste from ChatGPT or Claude, then hit the Publish button. Any of these sound familiar?
   
   👉 "Supercharge"
   👉 "Game-changer"
   👉 "Leverage"
   👉 "Synergy"
   👉 "Revolutionize"
   
   ... because let me tell ya, there's nothing revolutionary about these AI stock phrases.
   
   AI tends to rely on formulaic expressions because—as we know—its content is formed from regurgitated content that already exists. Basically, it loves the kinda of sh*t we used to say back in like, 2017.
   
   Look closely and you'll see that every sentence is structured in the exact same way, too. Not to mention AI's love of predictable transitions, including: "furthermore", "moreover", "additionally", and that final AI paragraph of "Remember: [*insert some repetitive tip here*]".
   
   We can do better than that, right? By spotting, then editing these to sound more natural and well... human, you will help the flow of what's written and how it's read, and ultimately the engagement.
   
   
   
   

2. Content quality indicators

   
   Let's split this AI detection technique into two parts: 1) General knowledge and 2) Human writing style.
   
   Firstly, you'll find that with most AI-generated content, it lacks expert knowledge. In fact, it was recently reported that LLMs provide "confidently wrong" answers to more than 60% of queries—ChatGPT Search was even found to cite the wrong article almost 40% of the time (!).
   
   There's a kinda surface-level treatment of complex topics, unless you really prod your AI tool to dig deeper OR you feed it with relevant, valuable, juicy data to help its output.
   
   The same goes for AI's understanding or translating into how human's actually speak or write—there's an absence of idiosyncratic writing style, personal anecdotes or unique perspectives, cultural references, and most of the time, AI really isn't funny in the slightest.
   
   AI-generated content can really miss the mark when it comes to facts and being completely up-to-date, as well as a human-level of authenticity and relatability. And sometimes, chat interfaces are kind enough to use a disclaimer saying it can make mistakes. By providing AI with facts, stats, and data (and editing the content it spits out to sound more like you or your brand), you'll see a huge difference in what your published content looks—and reads—like.
   
   

3. Structural red flags

   
   On a list of things that are incredibly robotic, I'd say #1 is being fully made out of aluminium or steel, and #2 is only providing content with uniform paragraph and sentence lengths.
   
   If you are guilty of AI-only content with ChatGPT, Claude, or equivalent, open up a few of them in different tabs, and play spot the difference with article structures.
   
   You'll find that:
   
   🤖 Paragraphs are all around the same length
   🤖 Each article has almost the exact same structure
   🤖 A lot of bullet point lists (the irony of this being a bullet point, I know)
   🤖 Absolutely no spontaneity in thought progression
   
   It doesn't have to be this way; your content does not have to be this sad.
   
   Whether you can take the time to train up your chosen AI tool to write more like you would (complete with training on brand messaging and brand training), you introduce AI agents to help you spot these AI 'faux-pas', or you simply make sure you check over the content yourself, you'll level up your content and see far more results.
   
   

4. AI-detection tools (that do it for you)

   
   If you're looking to see just how obvious it is that AI wrote your last blog or LinkedIn post, or if you're trying to suss out whether that freelance content writer is really worth the $$$, AI-detection tools are a pretty quick and easy way to spot AI-generated content.
   
   While they are not guaranteed to be 100% every time, it gives you a good idea of how original and authentic the content sounds. You know, the two traits that make content a lot more interesting to read.
   
   Here's some insight into how these tools work:
   
   ⭕ Perplexity analysis: Human writing typically shows higher perplexity (more unpredictable patterns) than AI-generated text
   ⭕ Burstiness measurement: Human writing tends to have 'bursts' of complexity followed by simpler passages—something that AI text can lack
   ⭕ Style and features: Examination of sentence structure, word choice, and punctuation usage that might reveal those common in AI writing.
   ⭕ Entropy patterns: The analysis of information density and distribution throughout the piece of text
   ⭕ Watermark detection: Some AI systems actually watermark their outputs with statistical patterns so detection tools can easily identify them
   
   And of course, for those interested, here are some AI-detection tool examples:
   
   👉 GPTZero
   👉 Originality.ai
   👉 Turnitin
   👉 Content at Scale's AI Detector
   👉 Writer.com's AI Content Detector

The moral of the story? Don't be lazy with AI content

The effectiveness of AI-generated content hinges on three critical factors: well-crafted prompts, the ability for your AI tool to intake specific brand information and guidelines, and vigilant human oversight. Both before and after the the initial content draft is complete, marketers need to remain actively involved in the content workflow to make sure it's ticking the quality and authenticity boxes. ✅

Now, I don't know if you've met her yet but Optimizely Opal is the ultimate AI sidekick in this content process. Integrating seamlessly at every stage of your content creation workflow, Opal helps ideate, create, and optimize, as well as enhance your speed, efficiency, and productivity. Fuel her up with good prompts and strategic direction, and you'll find the perfect balance between all this AI hype and human creativity.

As we always, always say: work smarter, not harder—and keep resonating with your audience.