AI Blindness: When Your Brain Learns to Stop Reading AI-Generated Content

Have you ever opened a document from a colleague, started reading, and felt your brain just... slide off the page? You are looking at the words, but the meaning does not register. You find yourself asking questions that were already answered in the text you supposedly read. If this sounds familiar, you might be experiencing what developer Rafal Cymerys calls "AI blindness" — and you are not alone.

Cymerys wrote a blog post titled "I'm becoming AI-blind" that struck a nerve across the tech community, racking up 366 points and 364 comments on Hacker News in under 24 hours. His thesis is simple and unsettling: after being bombarded with AI-generated content for the past few years, his brain has learned to automatically detect and ignore it.

What Is AI Blindness?

The term draws a direct parallel to "banner blindness" — the well-documented phenomenon where web users unconsciously ignore banner-style advertisements. After years of exposure, the brain learns to filter out anything that looks like an ad, skipping over it entirely without conscious effort. AI blindness works the same way, but for text.

Cymerys describes his experience:

"I've been catching myself having these little moments at work, when I'm trying to read a document someone has sent me and my brain somehow refuses to analyze it. It feels like I'm reading it, but I'm unable to focus on its content."

When he sat down to analyze these moments, he found a common denominator: the documents all showed strong traces of AI generation. A design document that read like a Claude copy-paste, complete with Claude-specific phrases. A 20-page marketing deck that mixed reasonable strategy with nonsensical technical gibberish. A requirements document that described a simple concept in verbose, meandering prose that sounded like an LLM's internal reasoning rather than a human's deliberate explanation.

The Telltale Signs of AI-Generated Text

One of the most interesting aspects of the Hacker News discussion was how many commenters independently described the same pattern recognition. The signals that trigger AI blindness are remarkably consistent:

• Overblown framing — describing mundane features as breakthroughs. If your RBAC checkbox configuration is pitched like the invention of fire, something is wrong.
• Specific LLM phrases and patterns — Claude's "This cuts just through it" or "The first gate is real," ChatGPT's "It's not just X, it's Y" construction, the ubiquitous "Let's dive in."
• Verbose explanations of simple concepts — using five paragraphs where two sentences would do, with hedging language that reveals uncertainty the author did not actually have.
• Mixing genuine insight with nonsense — a marketing strategy that makes sense paired with technical architecture claims that are clearly fabricated, as if the model filled in gaps it did not understand with confident-sounding gibberish.
• The overall flow — sentence structures that follow predictable patterns, paragraphs that all have the same shape, and a peculiar uniformity that no human writer would produce.

The Cost of AI Blindness

Here is where the story takes an ironic turn. The same AI that was supposed to make us more productive is now slowing us down in an entirely unexpected way. When your brain decides a document is AI-generated, you do not just skim it — you actively disengage from it. You stop absorbing information. You end up in prolonged back-and-forth conversations asking questions that were already answered, because you genuinely did not process the answers.

One Hacker News commenter captured the experience perfectly: "There's some psychological mechanism by which my brain immediately recognizes AI generated text and just short-circuits to 'there is no information here.' And when I force myself to read AI-generated text I realize I'm making my brain do creative work to impart meaning to the words. It is exhausting because my brain is literally trying to do a just-in-time rewrite of the text into something valuable."

This is the hidden tax of AI-generated content. It is not just that the content is lower quality. It is that encountering it actively degrades the reader's ability to process information. The brain spends energy detecting and filtering AI output, and then spends more energy trying to extract meaning from text that was never designed to convey meaning in the first place — it was designed to sound like it conveys meaning.

The Pre-Training Effect

Cymerys uses a telling phrase: he says he has been "pre-trained" on AI-generated content. After months or years of exposure to LinkedIn posts, emails, and websites that are full of text but empty of meaning, his brain learned to spot the patterns and automatically filter them out. This is not a conscious decision. It is an adaptive response, the same way humans learn to ignore billboards, skip TV commercials, or tune out background noise.

The implications are significant for anyone producing content in the AI era:

• Your AI-assisted document might be actively ignored, no matter how much effort you put into it. The reader's brain may decide it is not worth processing before they even get to your actual point.
• The problem compounds — every low-effort AI-generated document trains readers to be more skeptical of the next one, even if it is genuinely useful.
• Human-written content is becoming more valuable precisely because it is scarce. The signal-to-noise ratio of the internet is dropping, and anything that signals genuine human authorship cuts through.
• Trust is eroding at scale — when readers cannot distinguish good AI output from bad AI output, they treat all AI output as suspect.

Even AI Images Are Not Safe

The AI blindness effect extends beyond text. Cymerys shares a personal anecdote from a vacation on the Baltic coast. He was hungry, walking past restaurants, and saw a photo of food in one restaurant window that his brain immediately flagged as AI-generated and ignored. A minute later, something made him walk back. On closer inspection, the photo was real — it was a photo of quiche with what appeared to be mold. His AI blindness had caused him to nearly skip a restaurant based on a real photo that his brain had categorized as AI slop.

This is the dangerous flip side of AI blindness: false positives. As our brains become more aggressive at filtering AI-generated content, we risk filtering out genuine human content that happens to share surface-level characteristics with AI output. A poorly written but sincere email gets dismissed. A real photo with unusual lighting gets flagged as synthetic. The filter that protects us from noise also cuts us off from signal.

What This Means for Communication

If you work in any role that involves producing written communication — and in 2026, that is most roles — AI blindness has practical implications for how you should work:

• Edit aggressively — if you use AI to draft, rewrite every paragraph in your own voice. The first draft from an LLM is a starting point, not a finished product. If it sounds like AI, your reader's brain will discard it.
• Be concise — one of the strongest AI signals is verbosity. Say what you mean in as few words as possible. Brevity signals human authorship because it requires understanding what to leave out.
• Include specific details — AI tends to generalize. Concrete numbers, named individuals, and specific project details are signals that a human wrote the document.
• Strip the LLM tics — remove "delve into," "landscape," "in the realm of," and the construction "It's not just X, it's Y." These phrases are instant AI signals.
• Show your reasoning — AI-generated documents present conclusions without showing the thinking behind them. If you made a decision, explain why. The reasoning process is uniquely human and hard to fake.

The Broader Implications

AI blindness is more than a personal productivity quirk. It represents a fundamental shift in how humans interact with written information. For the first time in history, we are developing automatic filters not for advertisements or spam, but for the style and substance of communication itself. We are learning to distrust fluency.

This has profound implications for the AI industry. Companies like Anthropic are rolling out text watermarking to comply with EU regulations. But watermarking may be beside the point — humans are already developing their own watermarking system, built into the pre-conscious processing of their brains. The question is not whether machines can detect AI text. It is whether humans will bother reading it.

The deeper lesson from Cymerys's post is not that AI-generated content is bad. It is that communication is a two-way street. The sender encodes meaning, and the receiver decodes it. When the sender uses a machine to encode, the receiver's brain may simply refuse to decode. Not out of protest or ideology, but out of the same adaptive efficiency that makes us skip banner ads without ever seeing them.

The AI industry loves to talk about scaling laws, context windows, and benchmark scores. But the most important metric may be one that no benchmark captures: whether a human brain, reading the output, decides it is worth the effort to pay attention. Increasingly, the answer is no — and that is a problem no amount of compute can solve.