How to recognize AI texts

Look at the picture. No doubt you’ll have noticed that the text there has been written by Artificial Intelligence (AI). The picture as such is human-made, however; made by me (with the help of Photoshop).
We can characterize 2024 in many ways, but we can say that 2024 was the year that AI made its definitive breakthrough. Of course, AI was already known among insiders and experts, and reports of its use had already reached the general public. Launched in 2022, in 2023 ChatGPT (to mention only one application of AI) had already gained over 100 million users, but since 2024 ChatGPT's website is among the 10 most-visited websites globally, and in 2025 it’s impact will only be stronger. AI has become inescapable.
AI can be very useful, but it can also be dangerous. As for the latter, often it generates inaccurate information, especially about recent “facts”. I noticed this myself when I asked ChatGPT to write a text about me, namely who I am as a philosopher. It was full of nonsense. It made me a professor in philosophy at a certain university (fake!); it told that I had written about certain subjects, which was also nonsense; etc. In fact, the whole text was nonsense. Moreover, it has also happened to me that people sent me mails written by AI, suggesting that they were written by real persons. AI can mislead also in other ways, and if you base your decisions on false or misleading AI generated texts, you can make enormous blunders. It can even be a matter of life and death, if you work in the healthcare sector. On the other hand, AI can be very useful, and it can help you save time and make you more productive. So, you certainly should not avoid using it. However, it is always important to know that a text you use has been written by AI, as just illustrated. Despite all its positive effects, “the rise of AI’s use has been matched by the rise in AI’s misuse to deceive, manipulate, and extort, creating the potential for significant economic and political implications”, as one website says.
If you have recognized the text in the blog photo as AI written, probably it was because of its woolly language. Nevertheless, AI texts are becoming more and more perfect, and maybe, in future, you cannot distinguish any longer AI generated texts from human written texts, or at least it will be very difficult. Even now, tests show that in many cases people see AI-generated texts as human written (and also the other way around). And some people simply have no sense of seeing that a text is AI generated. Of course, you can use an AI detector, but before you decide to use one you must have reason to do so. For in the end you cannot use an AI detector for every text you read. Therefore, here are some tips how to recognize AI texts without the use of such a detector.
As I read somewhere: “The magic – and danger – of [an AI text generator] lies in the illusion of correctness. The sentences they produce look right – they use the right kinds of words in the correct order.” And just this is also the weak point of AI generated texts, for such texts are often too beautiful, too correct. Such texts are often so beautifully and correctly written, that they become “woolly”, and that’s often a reason to become suspicious. But let me briefly list ten points that can help you recognize AI texts. (compiled from this website)
1) Repetitive writing: AI writing is characterized by repetitive phrases and ideas, while humans try to vary their words (I have even learned to do so at school, when writing texts).
2) AI texts are often formal or formulaic. It looks as if they come from a phrase book, and AI texts too often use common phrases and idioms.
3) Related to the former point is that AI texts frequently use certain words. Typical AI words are crucial, delve, dive, tapestry, furthermore, consequently, “not only but”, “in today’s adjective world of”. But there are more.
4) Presenting questionable facts and claims without giving sources. Don’t believe if you doubt and cannot check. (Should you actually always do)
5) Lack of personality. AI texts are monotonous, don’t use informal language, don’t use colloquialisms or slang, and don’t have a unique tone of voice.
6) AI texts are general and avoid details. They are not specific. Moreover, complex subjects are covered superficially, because AI lacks the depth and detail that come from first hand expertise.
7) The facts presented in an AI text are often outdated.
8) Absence of personal experiences, or they are formulated in a vague way.
9) Unconvincing storytelling or an incoherent narrative. No progression in a story and the story has no natural flow.
10) AI doesn’t understand sarcasm and the symbolic, non-literal meaning of words.
For this list, I used a blog by Satyo D., where you can find these points in more detail. Other websites list more or less the same characteristics of AI generated texts. Once you think that a text is AI generated, you can check it with an AI detector (there are many free detectors on the internet). However, also AI detectors can mislead, for also they sometimes make mistakes and don’t recognize AI written texts as written by AI or they think that human texts are AI generated. For aren’t there many people who use woolly language? Machines use woolly language because they don’t understand what they write, but isn’t it so that also humans often don’t understand what they are saying or writing? That they don’t understand themselves?
Happy New Year! (Note that this expression fits most of the ten points listed above. Nevertheless, it is not AI generated.)