Why AI texts annoy readers: a Habr author analyzed reactions to AI writing style

A Habr author published an analysis of why materials created or polished by neural networks so often annoy readers. The author tested this himself: he published AI-generated texts on Habr, tracked how people reacted to the presentation, rhythm, and overall tone, and concluded that the audience usually catches not a substitution of ideas, but a substitution of human form.

Self-Experiment

The author describes an almost laboratory test. He intentionally posted materials on Habr gathered with the help of neural networks and tracked how people reacted to the delivery, rhythm, and overall tone. The most telling moment came when he took his own old article with a good rating, "polished" it a bit with AI, and gave it to the audience again.

The ideas, arguments, and author remained the same, but the reaction sharply deteriorated. This gave him a clean comparison of reactions to content versus packaging. Instead of recognizing a familiar style, readers saw what is often called "generatedness": too even a pace, clean transitions between blocks, and safe formulations without living irregularity.

The experiment proved important precisely because it eliminates the main counterargument — it's not about a weak topic or a new author. The shell changes, and with it, trust changes. This became the main observation of the whole story for the author.

"Same author, same thoughts. Only the form changed."

What Reveals AI

The text connects this irritation not only to the cultural trend of AI criticism, but also to basic psychology. The reader constantly evaluates how much the speech resembles a signal from a living interlocutor: where there is risk, where personal choice is heard, where strange but meaningful turns are noticeable. When everything is too averaged and smooth, the brain begins to perceive the material as synthetic, even if a human actually wrote it or at least edited it heavily. Suspicion is usually triggered by such signs:

• perfectly even rhythm of phrases without natural jumps
• template connectors between paragraphs and predictable conclusions
• generalizations instead of observations that can be verified
• repetition of one thought in different words
• sterile tone without authorial risk and roughness

At the same time, the author emphasizes an important issue: such intuition often makes mistakes. A human can also write dryly, by template, or after heavy editing. So the "AI detector" in the head works not as an analyzer of text origin, but as an alarm signal: we are dealing with speech that has little individuality and too much statistical average. Hence the false alarms and aggression in the comments. This is precisely why feelings alone are no longer sufficient for assessment.

What's Wrong with Code

This same logic transfers to programming. Code from neural networks often looks neat: it has clear structure, correct variable names, familiar patterns, and even decent comments. But this very external well-being can be a trap.

The model reproduces the average image of a "good solution" well, but is weaker at holding real project constraints: architectural history, non-obvious invariants, fragile integration points, and edge cases. In real development, this usually decides the fate of the result. Because of this, the developer faces a special kind of distrust.

The error is not always noticeable immediately because the code looks convincing and reads easily. The problem surfaces later — on unusual data, in combination with neighboring modules, or when trying to maintain the solution further. Essentially, text and code behave the same way here: the neural network often produces a plausible form faster than deep understanding of context.

And when a person notices this gap, irritation only intensifies.

What This Means

The main conclusion is not that AI content should be banned, but that it cannot be evaluated by surface alone. For authors, this is a signal to polish text less into impersonal smoothness; for editors, to preserve voice and specificity; and for developers, to check neural network code as a draft, not as a finished solution. In the coming years, those who will win are not those who simply connected AI, but those who learned not to lose human irregularity and context.