Why AI Can't Write Like Humans: The Authenticity Problem

Key Insights

• AI models each have distinct "voices" — Gemini is optimistic, Claude is sharp but measured, and OpenAI's systems are dispassionate — making consistent writing tone nearly impossible when using multiple AI tools together
• Perfect isn't authentic — the imperfections in human writing (vinyl pops, film grain, odd punctuation, unexpected phrasing) create genuine connection that AI-generated text fundamentally lacks
• An "AI editorial council" fails because multiple AI models cannot reach consensus on tone, style, or direction, resulting in disjointed, contradictory revisions rather than cohesive improvement
• Authenticity requires imperfection — AI's ability to generate flawless documentation doesn't translate to meaningful, resonant writing that readers actually connect with
• The future of writing isn't AI replacement — it's understanding that human imperfection, vulnerability, and unique voice are what separate good writing from technically perfect but soulless text

The AI Voice Problem: Why Multiple Models Create Chaos

The challenge with AI-assisted writing becomes immediately apparent when you try something intuitive: using multiple AI systems to improve your work. After 16 years of writing, testing more than 10 different AI systems, and experimenting with fine-tuned models designed to mimic personal voice, the harsh truth emerged — AI doesn't solve the writing problem. It multiplies it.

Each major AI model brings its own distinct personality to the table. Gemini tends toward optimistic, sunshine-filled language. Claude delivers sharp insights wrapped in a languid, thoughtful tone. OpenAI's systems operate with clinical detachment, prioritizing information over emotion. These aren't bugs in the system; they're fundamental characteristics baked into how each model was trained and designed.

When you ask three different AI models to edit the same piece, you're not getting three perspectives on the same vision. You're getting three completely different visions competing for dominance. It's like hiring three editors with fundamentally incompatible philosophies and asking them to work on the same manuscript simultaneously. The result isn't an elegant mosaic of insights. It's a fingerpaint disaster — each model insisting its approach is correct while undermining the others' contributions.

This problem intensifies because AI models are remarkably confident in their assessments. They don't hesitate or suggest compromise. They deliver criticism directly, casually cruel in their pursuit of "improvement," offering verdicts like: "This is a three-beer conversation mistaken for a finished essay. Pick one angle and develop it fully." Multiply that feedback by three competing AI systems, and you've created a cacophony of contradictory demands rather than a coherent editorial strategy.

The Authenticity Crisis: Why Perfection Feels Empty

Here's the uncomfortable truth that AI forces us to confront: perfect writing is often lifeless writing. The pops and crackles of vinyl records, the slight color shifts and grain in Kodachrome film, the awkward analogy that almost works but not quite — these imperfections are precisely what make analog media feel authentic and alive.

The same principle applies to human writing. The idiosyncratic punctuation choices, the unconventional phrasing, the specific metaphors that emerge from lived experience rather than algorithmic optimization — these are the signatures of authentic voice. They're also the things AI struggles to produce because authenticity isn't a feature that can be engineered. It's a byproduct of genuine human limitation, perspective, and vulnerability.

Consider what makes writing truly memorable. It's rarely the technically perfect sentence. It's the unexpected turn of phrase that catches you off guard. It's the vulnerability that shows through imperfect word choice. It's the specific detail that reveals the author's particular way of seeing the world. These elements exist precisely because a human made specific choices — sometimes odd, sometimes risky, always revealing something true about their perspective.

AI, by contrast, optimizes toward the safe, the conventional, the statistically likely. It smooths away rough edges and quirks in pursuit of clarity and polish. The result reads like it was generated by algorithm, because it was. Even the most sophisticated models produce output that feels slightly artificial, slightly off in ways readers unconsciously detect even if they can't articulate why.

Building Better Writing: The Role of Imperfection and Voice

The real revelation from extensive experimentation with AI writing tools comes down to this: the goal isn't to create AI that writes like humans. The goal is to understand what makes human writing valuable in the first place, then protect those elements while selectively automating what's genuinely tedious.

For years, the writing advice industry has pushed toward perfection. Remove the adverbs. Eliminate passive voice. Make every sentence lean and efficient. This advice has merit for clarity and readability. But taken to its extreme, it produces writing that's technically proficient but emotionally hollow. It reads like it was generated by someone following a checklist rather than someone with something genuine to say.

The writers whose work endures are rarely the ones who mastered perfect technique. They're the ones who developed a strong, recognizable voice — often through countless failures, public embarrassments, and willingness to break the "rules" when it served their purpose. They're the ones whose imperfections became part of their signature.

This matters for how we approach AI in writing. Rather than using AI to move toward generic perfection, the smarter approach is using AI to handle the genuinely tedious parts (research organization, first-draft scaffolding, copy editing mechanics) while protecting the irreplaceable human contributions: voice, perspective, risk-taking, and the specific imperfections that reveal character.

Why AI Documentation Will Excel While AI Literature Will Always Fall Short

There's an important distinction worth making: AI is genuinely excellent at certain writing tasks. Technical documentation, instructional manuals, structured guides, and repetitive business communication are all areas where AI's strengths — consistency, comprehensiveness, absence of ego — become genuine advantages. The world needs better documentation, and AI can deliver it at scale.

But there's a ceiling to what AI can accomplish with more expressive, resonant writing. The ability to synthesize images, generate video, and produce reams of text doesn't solve the core problem: creating something that feels alive requires imperfection, risk, and genuine expression. You can't algorithmically generate authenticity.

This is why the writers who will thrive in the AI era aren't those trying to write like AI or compete with AI on AI's terms. They're the ones leaning deeper into their unique voice, their particular perspective, their willingness to be imperfect and specific and vulnerable. These are the things AI can't replicate because they emerge from genuine human limitation and choice.

The future of writing isn't a conflict between humans and AI systems. It's a question of what we actually value in writing and which aspects of the writing process deserve human judgment. Spoiler alert: it's not the polish. It's the voice, the vision, the imperfect authenticity that makes something worth reading in the first place.

Conclusion

After testing a dozen AI systems, hiring editors, and experimenting with multiple approaches to AI-assisted writing, the lesson is clear: imperfection is the signature of authenticity. AI will get better at many aspects of writing. But it will always struggle with the elements that matter most — genuine voice, meaningful risk-taking, and the specific imperfections that reveal something true about the writer. The writers who win aren't those who use AI to approach perfect technical proficiency. They're the ones who protect their imperfect humanity while strategically leveraging AI for everything else. In writing, as in life, the flaws are often the most valuable parts.

---

Original source: Observations on Writing with AI

powered by osmu.app