Be Your Own AI Early Warning System

24 Oct, 2025

Recently, my colleague wanted to know what I thought of a Viva Engage post by another colleague.

The content was not trivial; it dealt with a complex definition of terms, contextualization, and the impact on many areas of our work. It was heavy stuff, so to speak, aimed at a wide audience, including management. Not unusual for the colleague who wrote it.

Although I hadn't noticed it at first, it dawned on me after reading it several times that the article must have been written entirely or largely by AI. I was too busy analyzing the content in detail to see the elephant in the room, or in this case, in the article.

Once the switch was flipped, I intuitively thought first of familiar, individual characteristics that are typical of AI. For example, the discovery of “em dashes” is very popular, to the point of ridiculousness¹, in this type of exercise. More obvious is the extensive use of lists, numbering, and the like. All of this was also included in the article. But another characteristic suddenly stood out or came to my attention: I found it incredibly difficult to follow the sentences. The sentences were, of course, grammatically flawless, the choice of words was okay-ish, sometimes even creative, and there was a lot going on on the (digital) paper.

But still — I had to read each sentence very carefully and repeatedly in order to be able to follow the content to some extent. Even though individual paragraphs remained true to a particular style, there was no common thread, no coherence between them, and the “big picture” somehow didn't fit together. Not necessarily because of the content (which in this case was already dubious enough even without AI), the meaning, or the stylistic choices. No, what was missing was the “invisible hand” of the author, reaching out to me as a reader and guiding me through the train of thought — however chaotic and bumpy it may be. This kind of accompaniment, almost intimate participation in what is going on in the author's mind, is so inherently human that its absence is somehow disturbing. And I felt that, at first only subconsciously, but now also obviously.

And there, once again, a circle had closed for me, because I had had exactly this feeling again and again recently when I had tried to read texts by colleagues, from the internet, or from the ether in general. At first, it was just a realization of the day (TIL!), but I had the feeling that I had found something worth observing further.

So whenever I lose track after just a few words or sentences, often having to “rewind” while reading, my AI early warning system will now kick in and, together with the other clues, I will expose the authorship of the text!

Of course, there will also be a 'false positive' rate. For me, when these phenomena occur as a reader, it simply means that the author is not good. Ultimately, then, it doesn't matter whether the author is human or the output of a computer program.

In fact, I'm not sure how many new writers would feel offended or even honored when they are equated with generative AI. Now that, thanks to AI, everyone who previously claimed that writing wasn't their thing can suddenly write texts diligently and unabashedly, the number of “writers” has skyrocketed. But it's also true that the wheat is likely to be separated from the chaff even more clearly. It's an incentive to look closely, or rather read closely, and check the quality of texts from the outset, rather than just rushing into the content, as I did at the beginning.

Anyway, I'm looking forward to my new early warning system and am curious to see whether it will stand the test of time, whether AI will simply get better here too, or whether it will all turn out to be a big deception.