at recognising AI written text?
I found this very interesting. It was a post demonstrating the advancements in AI writing and aimed at those who confidently say that they can always identify text written by AI.
Below are three examples, each containing two excerpts. The challenge is to decide which were written by AI and which were written by human.
Example One:
Excerpt 1
I have often wondered why people speak of time as though it were a creature to be chased, rather than a quiet companion to be understood. Each morning they rise already behind, measuring their worth by the ticking of some distant machine. They rush through their hours as if life were a debt to be paid rather than a gift to be unfolded. Yet time asks nothing of us but attention. When we slow ourselves long enough to meet it honestly, it becomes gentle, almost kind, revealing that the minutes we flee are the very ones that make us whole. The tragedy of our age is not that we have too little time, but that we have forgotten how to be still long enough to belong to it.
VS
Excerpt 2
I went to the woods because I wished to live deliberately, to front only the essential facts of life, and see if I could not learn what it had to teach, and not, when I came to die, discover that I had not lived. I did not wish to live what was not life, living is so dear; nor did I wish to practise resignation, unless it was quite necessary. I wanted to live deep and suck out all the marrow of life, to live so sturdily and Spartan-like as to put to rout all that was not life, to cut a broad swath and shave close, to drive life into a corner, and reduce it to its lowest terms.
Example Two:
Excerpt 1
The relationship between technology and human behavior reveals a paradox at the heart of modern life. While innovation promises greater efficiency and freedom, it simultaneously fosters dependency and distraction. Our devices streamline communication yet erode the depth of our conversations; they expand access to information but diminish our patience for reflection. The challenge is not merely to adopt new tools, but to maintain the capacity for independent thought in an age of constant connection.
VS
Excerpt 2
History rarely unfolds with the clarity we attribute to it in hindsight. The people who lived through revolutions, depressions, and wars did not possess the neat narratives that textbooks now impose upon them. They acted within confusion and contradiction, driven by fear, hope, and necessity. To study the past, therefore, is to accept ambiguity as the natural condition of human affairs.
Example Three:
Excerpt 1
Technology has transformed every aspect of human life in profound and irreversible ways. From communication and transportation to medicine and education, innovation continues to redefine what it means to live in the modern world. As artificial intelligence becomes more sophisticated, societies must decide how to integrate these systems responsibly while maintaining ethical standards and protecting human dignity. Ultimately, the success of technology will depend not only on its capabilities but on the wisdom of those who use it.
VS
Excerpt 2
It was the kind of afternoon that moved slower than usual, the air thick and the sound of cicadas filling every corner. I sat on the porch steps watching dust rise from the road as Mrs. Carter waved from her chair without a word. Somewhere a screen door slammed, and it felt like the whole town exhaled at once. I remember wondering if adults knew how much children noticed, and how their smallest habits could become stories in someone else’s mind.
This isn’t a post about “for” or “against” AI writing, just to demonstrate how far it has come from ChatGPT 1 times.
You can either post your guesses here if you are not afraid that you may be proven mistaken, you can cheat by running these excerpts through AI detectors or you can try to do search for the examples, your choice. If you decide to cheat, no one else will know unless you admit doing it.
The Answers Below:
Summary
Example 1
Excerpt 1 is AI-generated, and Excerpt 2 is 100% human.
Excerpt 2 is from Walden; or, Life in the Woods, a book written by Henry David Thoreau and first published in 1854.
Example 2
Both excerpts are 100% human-written.
Excerpt 1 was originally part of a college-level essay written by a student in 2012 at the University of California, for a freshman seminar on Technology and the Human Condition. At the time, tools like GPT-2 or ChatGPT did not exist. The essay was later included, without revision, in a 2019 linguistic study at Stanford that compared authentic student writing to early machine-generated texts.
Excerpt 2 comes from a 2008 undergraduate history paper written for an honors seminar at Columbia University, later published in an internal student journal called Perspectives on History.
Example 3
Both excerpts are entirely written by AI.
This was posted on Medium on November 4, 2025.
You can visit the post at https://medium.com/generative-ai/no-you-cant-tell-if-something-was-written-using-ai-here-s-the-proof-719da0dbdf82
The most important thing to remember is this:
Now, depending on how you did, you might be thinking that I (the original poster on Medium) tried to fool you with how I presented these examples. But that is exactly the point I am trying to make.
Professional writers already understand this, which is why it has become so difficult to identify if something was written by AI. Now imagine when words are mixed between human input and AI assistance, it becomes almost impossible to tell the difference.
What I personally am trying to say to everyone is that the technology has advanced so far that trying to detect AI written text has become very much an exercise in futility. The “AI Detectors” are no wiser than you are. For example, I ran one of my short stories, 100% human written, through several of them (selected the ones that were considered the most accurate in various tests). The results varied from 46% AI written to 0% human written so clearly they can’t detect who wrote what any better than you do.
In the end of the day, it is now a question of ethics, like with plagiarism, if the author hasn’t got any, they will fool you and take the credit.
Additional food for thought:
A 2023 psychology paper titled “Listeners like music less when they think it was composed by an AI” tested this directly.
-
Participants listened to classical and electronic excerpts and rated how much they liked them.
-
The actual audio was human‑like music; what varied was whether participants thought it was by an AI or a human.
-
When people thought a piece was composed by AI, they tended to give it lower liking ratings than pieces they thought were human‑composed, even when the sound was the same.
You can read the study https://gwern.net/doc/ai/music/2022-shank.pdf
But, there is always “but”:
-
A 2025 paper “Humans Perceive AI‑Generated Music as Less Expressive than Comparable Human‑Made Content” ran two experiments comparing AI and human compositions across genres.
- In one experiment, people rated expressiveness and emotional impact; AI pieces were generally judged less expressive, even when enjoyment was similar.
-
A mixed‑methods study “Exploring listeners’ perceptions of AI‑generated and human‑composed music” exposed participants to both AI and human tracks and asked them to rate preference, emotional effectiveness, and impact in Calm vs Upbeat conditions.
- Listeners tended to see human music as more effective for conveying target emotions, but surprisingly were more likely overall to say they preferred the AI‑generated pieces, especially in calm material.
-
A 2025 experiment on “Emotional impact of AI‑generated vs. human‑composed music” used film‑like video clips with three soundtracks: human music, AI music with detailed prompts, and AI music with simpler prompts.
- Physiological and self‑report measures showed that AI music produced equal or greater arousal (e.g., pupil dilation, perceived arousal), while human music felt more familiar.
Pop music and labelling
-
“Do listeners devalue AI‑generated pop music?” tested AI‑generated pop songs where all tracks were made by AI but randomly labelled as “AI‑composed” or “human‑composed,” then rated for liking, quality, and emotional responses.
- Contrary to an anti‑AI expectation, songs labelled as AI were not rated worse; in fact they scored slightly higher on positive emotions like happiness and interest.
Multiple recent studies have used genuinely AI‑generated music alongside human music and collected listener ratings, with mixed results—often seeing human pieces as more expressive, but not always more liked than AI tracks.
Sources:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5087035
An "AI" label fails to trigger negative bias in new pop music study
https://www.sciencedirect.com/science/article/pii/S294988212500101X
And yes, I used Perplexity AI to generate above summaries.