This Blogpost Was NOT Written by AI

It was there again – in that blogpost I was reading. The telltale signs of AI authorship. Those signs seem to appear in most blogposts these days. You can identify them because they tend to make blog content bland. Sterilized. Even predictable.

AI-generated content has become standard fare for online publishing. It’s easy to produce. It is generally accurate. And it doesn’t require the author to spend much time pushing out content (and then taking credit for it).

So, how do you know if the blog you are reading is original content or something aggregated and published by AI?

The Telltale Signs AI-generated Blog Content

You can almost bet that any AI-generated blogpost you read will have a rigid, almost mechanical structure to it. The blog intro will preview exactly 3-5 points, followed by body copy that addresses each point in the same order – with generally the same paragraph length – and a conclusion that basically restates the introduction. Contrast that with blogs written by real humans. They are not so uniform. Some points may require multiple paragraphs while others only need one sentence.

AI-generated content is pretty stiff. It holds tightly on-topic. Conversely, human writers tend to wander, occasionally with some content that may not perfectly serve the blog’s thesis.

Another difference between AI and human-generated content is that while AI content may touch on many subtopics, it rarely goes deep on any one of them. After all, AI doesn’t have any genuine curiosity. It’s just delivering data; its purpose is to be comprehensive.

No Point-of-View

Unlike human writers, AI does not address a subject from an embedded point-of-view. It doesn’t have an opinion about anything. No real-life experience. No admission of uncertainly, like “I tried this and failed spectacularly.” Instead, it presents generic claims with confident neutrality.

More Signs of AI-generation

Beyond the structural aspects of blogposts, here are some additional signs that indicate the blog you are reading is a product of artificial intelligence:

• Perfectly balanced sentence lengths. Are all the sentences moderate-length and evenly paced throughout the article? That’s not natural human writing. We humans tend to write with a more erratic rhythm: a short sentence followed by a longer sentence; another short sentence; then possibly a medium-length sentence; and so on.
• Formulaic transitions – such as ”Moreover,” “Furthermore,” “Additionally,” or “That said” – used mechanically at the beginning of a paragraph rather than because logic actually calls for it to be there.
• The “it’s-not-just-X,-it’s-Y” construction, such as “This isn’t just a tool, it’s a revolution…” This construction shows up constantly in AI copy.
• “Triplet” lists. This is where you are served up things clustered in groups of three (“fast, reliable and scalable”). This stylistic construction appears much more often in AI writing than in natural human copy construction.
• Overused vocabulary that frequently shows up in AI-generated text, including words and phrases like “delve;” “tapestry;” “navigate;” “elevate;” “seamless;” “game-changer;” “in today’s fast-paced world;” and “in the ever-evolving…” You get the picture.
• Absence of the use of contractions.
• Absence of concrete, verifiable specifics, citing of vague examples (“many businesses have found success…”) rather than using a named company, date, number and/or anecdote. And when numbers do appear, they are often suspiciously rounded or unsourced.

Machines Produce Mechanical Output

Granted, none of these signs are proof that a blogpost was AI-generated. However, when you see a combination of several of these telltale signs of AI-generation stacking up together – especially the structural ones – there’s a good chance you are not reading human-made content. It doesn’t mean the content is wrong, it’s just not created by a real person. And it’s probably not as interesting as content created by a real person.

You are probably aware there are accuracy problems with AI called “hallucinations.” AI models often confidently state outdated information, invent statistics, or misattribute quotes. Furthermore, AI content constantly struggles with current events and fast-moving topics. And because it is machine-generated, AI doesn’t know what it doesn’t know…

As a result, readers are beginning to recognize AI-written posts. Once readers suspect a post is AI-generated, they tend to discount its authority.

The Value of Human-generated Content

AI cannot replace human experience and judgment about what matters. A human writer decides what to include based on genuine curiosity, professional expertise, personal experience, or at least a sense of what his specific readership needs or wants – not a statistical average about “what is usually said about this topic.”

AI content lacks two things that make writing trustworthy: 1) someone who actually knows, and 2) someone who is accountable for being right!

Will this change to a point where AI can fully replace human-written content as it continues to evolve? I wouldn’t bet against it.

A Confession

I lied at the beginning of this blog. Artificial intelligence was used to help write it…