Chatscape

I work in tech. I have many thoughts about AI. I wanted a place to put them.

#AI-settings #expectations #prompting #working-with-AI #writing

20 May, 2026 elevating the flaws of humanity to beat the machine
People are openly admitting to "dumbing down" their writing so as not to be mistaken for using AI. I couldn't find a word for this. When I tried to look it up "dumbwriting" came up as a result but if you look into dumbwriting, that's more about embracing messiness/not seeking perfection, and less about being mistaken for AI.

However, what I did find was multiple posts, articles, Reddit conversations, and other thoughtful pieces across a wide array of sites about how/why people are doing this. It's not just happening in the classroom. People do it in Reddit comments, on their blogs, and one of my penpals admitted to doing this in typed letters before she caught herself.

What am I talking about? It's when someone intentionally leaves (or adds) typos, run-on sentences, or grammatical errors in their "finished" writing. Why do people do it outside of academic contexts? To "beat" the AI. Look at how messy my writing is! Clearly, it's not written by an LLM.

I have done it too. It's easy to fall into this idea/trend that it's the "best way" to fight against polished AI output. But once I notice I do something new, I tend to research into WHY - why I do it, why others do it, where the compulsion comes from - and most of the time I end up stopping as a result of what I find. This is one of those times.

Look, I want to be clear before I dive into this perspective. I write for a living so I cannot afford to intentionally dumb down my professional work, and I'm now hard pressed to do it in my personal writing too.

I like messy. In no way am I claiming in this article that people should strive for perfection. In fact, exploring messy formatting and going off the rules is an excellent way to find or form your creative voice. Sometimes the messy fits, and you keep doing it forever; sometimes the contrast and helps you find a different path. I am a champion of "trying new things" and "bending the rules" for creative purposes.

As a strategy to "beat AI," though, choosing to dumb down work that's not graded on AI-checker output is futile (that nuance is addressed later). It misunderstands how AI works and underestimates the depth of what humans can create. If the primary difference between human expression and machine output comes down to spelling errors and lowercase letters, that's a low bar for us.

This isn't Our First Technology Tantrum
This isn't a unique reaction to new technology. Almost every major technological leap triggers a counter-cultural retreat into innefficiency. Here are two examples:
- The Industrial Revolution brought a sudden spike in the value of slow, flawed, handmade goods the second factories took over. Suddenly a handmade chair with all its flaws was valued higher than a factory-made one (as, I think, it should be - but this change in value didn't come out of nowhere). Previous to the technology capable of creating them in the mid-1800s, all chairs were handmade; individual artists and designers may have had higher priced chairs, but there was no strict division between "factory-made" vs "handmade" until factories arrived. They argued that machine-made goods were soulless and stripped of humanity.
- The story of the calculator reminds me a lot of our arguments surrounding AI. When calculators became commonplace in the 1970s/80s, there was a sudden obsession in the educational community with long-hand math as a moral imperative when machines could do arithmetic for us. Certainly, a calculator is a different type of technology than AI, but the same "cognitive decline" arguments came out and as a result for decades, kids couldn't even have a calculator in math class without getting in trouble. For ages the main argument was that "you won't always have a calculator in your pocket," but we do now. The National Council of Teachers of Mathematics (NCTM) pushed for a cultural shift on calculators, convincing educators that outsourcing raw arithmetic freed up cognitive space for higher-level conceptual math.
The interesting part the reaction to AI for me is that usually the human pushback involves lifting ourselves higher: it emphasizes high effort and complex intellectual challenges. It requires you to use your brain. This is different. We're embracing "shitpost aesthetic" in our writing. This is a deliberate engagement in anti-perfection against sterile AI ouptut. We are intentionally "trashing" our art to attempt to create something AI cannot mimic. I hestitate to call this "anti-intellectualism," because I don't think it is: I think it's specifically a middle-finger pointed at the corporate optimization of AI. And that's awesome. But I'm not sure it's the best way to do it.

AI is Our Baby
We act like LLMs are cold, flawless machines, but they're actually just a baby created by all of humanity. AI datasets are a mirror of us. There are so many inconsistencies in the idea that AI = perfection, humans = flawed, and that's the difference between us. (And frankly, contradictory: people complain about AI's flaws all the time!)

AI quirks are not anomalies; they're often human flaws that have been amplified and sped up by automation. This is long and there's more to go, so I'll stick to two examples, but there are many more ways that AI mirrors humans.
- We view AI "hallucinations" as a machine flaw, but humans are constantly assuming things. We fill in the blanks with biased or incorrect data all the time. "Hallucinations" aren't an error, they're a feature of predictive language. They look like "flaws" only when you're looking through the lens of "machine perfection." There's no database of factual truths; like humans, AI guesses based on context. Expecting perfection or objective truth from AI is the flaw in this equation.
- AI inherits our worst traits from its training data: it's very confident even when it's wrong, and it carries all of our structural biases. Have you ever been on Reddit and seen a thread where someone pointed out the Dunning-Kruger effect about a person in the replies. AI does that at a faster pace. It is trained on human text, and copies authoritative human writing styles. If you're interested in reading more, a 2025 study maps out how these systems become aggressively confident when they're wrong.
Sabotaging Your Writing Intentionally is a Trap
To be clear, if you don't want to write with AI, don't. And I'll repeat: Experimenting with raw and weird styles for the exploratory joy of finding your voice is great... but doing it solely out of fear so you won't be mistaken for an LLM is a trap.

Good writing can be good writing, too. If a human writes a beautifully structured, clear, and articulate piece, it's still good human writing—even if a lazy reader glances at its clean formatting and thinks "looks like AI." I will not be letting AI dictate the limits of my excellence.

There's an accessibility tax. When you intentionally break your prose to make a point about your humanity, you lose clarity. You're turning away or frustrating other people including non-native speakers, individuals with disabilities, and anyone relying on screen reader software. Some people only write for themselves, maybe this isn't a concern for you.

You will hit a dead end. What happens when LLMs inevitably become more natural and "human-like" in their output? What happens when people get better at prompting? Will you stop writing?

AI can imitate messy human writing. The core argument for doing this is flawed. AI can imitate a messy, typo-ridden post if prompted well. Most AI output looks identical right now because the average person doesn't know how to prompt for uniqueness or variability.

...And the Exception (Institutional Failure)
There is one major exception where this kind of writing is necessary, not on fault of the writer. In academic contexts, the reality is that students are forced to change their natural voice because AI detectors associate smooth, varied, and competent grammar with machine generation. To avoid being falsely accused, students are practicing "strategic incompetence" and intentionally flattening their assignments, discussions, or other writing. They know how to write a flawless statement, but deliberately choose to write a worse version because otherwise it will be marked as "suspicious."

If you are a student trapped in an academic system that relies on automated AI-checkers, you're in a tight spot. If you write "too well," the algorithm may flag you, and your institution likely hasn't built the human touchpoints required to protect you. In an ecosystem like that, strategic incompetence is necessary... but it's also a symptom of systemic failure. (I'll post about this thought another time.)

What the Machine Actually Lacks
So this brings me back to the question I asked myself when confronting this habit: Outside of institutional necessity (school or work), why am I doing this to my writing? Why let the presence of this machine and how it's interpreted by others dictate the boundaries of my expression?

Instead of dropping our standards to meet a machine's current limitations, we should lean harder into what LLMs genuinely cannot replicate.
- Sensory Experiences: The literal, physical reality of navigating the world in your body.
- Personal Anecdotes: Specific, detailed memories, events that happened, people that exist.
- Non-linear Styles and Formats: Break the rules for structure and formatting in your expression. (Unless you're at school/work, then don't. Formatting is very important in some contexts.)
- Strong Opinions: AI loves to be everybody's friend and take both sides with balance. Heavy human conviction is difficult for a safety-tuned model to copy.
- Audience-Aware Language: As a tech writer using AI in my work it's clear to me that even if you provide knowledge about the audience, LLMs will never truly "understand" what they know, don't know, and the nuances of communicating with them. This can be true for blogs and informal writing too.
What else? I'm sure there are more. Don't shrink yourself in this fight. Make it weird and more specific, but keep it yours.

(Can you tell who wrote this post? I left all my mistakes in.)

Further Reading
PS: A recent post I made on not getting left behind by AI was featured on Technically Good, which is very exciting for me because that is one of the blogs that inspired me to start this one.

That post awarded me my first time on the Bearblog Trending feed which made me nervous and vulnerable, but excited to get my thoughts out there for people to agree or disagree.

PPS: I updated my About page to add some more context and to add my AI transparency statement.

This isn't Our First Technology Tantrum

AI is Our Baby

Sabotaging Your Writing Intentionally is a Trap

...And the Exception (Institutional Failure)

What the Machine Actually Lacks

Further Reading

See all →