Bad and Good Side of AI Influence on Written Language
There’s a growing wave of articles and posts warning us that ChatGPT is flattening our writing, influencing language, and standardizing communication (e.g. this of Raffaello Castellano). I recently came across an article on The Verge, also cited by Castellano, that discusses recent academic studies:
“Empirical evidence of Large Language Model’s influence on human spoken communication” ( Hiromu Yakura et al., 2024). https://arxiv.org/abs/2409.01754
“Mapping the Increasing Use of LLMs in Scientific Papers” ( Weixin Liang et al., 2024) https://arxiv.org/abs/2404.01268
The papers analyzes measurable linguistic shifts caused by language models like ChatGPT and highlights how human-written texts are beginning to resemble AI outputs, particularly in word choice and sentence structure.
While the concern is understandable, I believe the framing is misleading.
Yes, ChatGPT, and generative AI in general, does influence writing style. That’s inevitable. But so did:
- The globalization of internet language
- The anglicization of local expressions
- The absorption of foreign terms into daily use
- The memefication of culture, (https://www.linkedin.com/pulse/heres-johnny-meme-mythology-marketing-copyright-lunardi-papasogli-1144f)
- The television era, which condensed speech for mass consumption (Neuman, 1995)
- YouTube, TikTok, and Twitter, which rewired our rhythm of storytelling and attention span
Every major medium shapes the way we speak and write. ChatGPT is not the first influence, and it certainly won’t be the last.
🔄 The Real Problem: Outsourcing Thought, Not Just Writing
If your language is becoming flat or predictable because of AI, the issue is not the tool, it’s how you’re using it.
Blaming ChatGPT for generic writing is like blaming a calculator for poor math skills. If you just let the machine do all the work, you’re missing the point.
The danger isn’t in using AI. It’s in not collaborating with it.
A healthy interaction with ChatGPT means:
- Prompting it with nuance
- Refining its output with your voice
- Challenging its suggestions
- Using it as a creative partner, not a crutch
Treating AI like a ghostwriter leads to exactly the same outcome as students memorizing lessons without thinking: repetition without comprehension.
📚 This Is an Educational Challenge, Not a Technological One
What we’re really seeing is not a linguistic crisis, but a critical thinking deficit. And the solution isn’t banning AI, it’s promoting:
- Media literacy
- Cross-referencing multiple sources (starting with books!)
- Curiosity about language and form
- Personalized thinking, not templated content
We’re not facing a future where AI erases human voice. We’re facing a present where too many people are willing to give up theirs.
💡 Language Evolves. But Responsibility Stays Human.
Let’s stop blaming the tools for the choices we make.
Language always adapts to its environment. The printing press standardized spelling. TV and radio changed sentence structure. Social media brought emojis and hashtags into the mix. And now, generative AI offers another layer.
We’re not being forced to sound like ChatGPT, we’re choosing to, often because it’s fast, safe, and good enough.
But good enough is not your voice. It’s not insight. And it’s not communication at its best.
So yes, maybe AI made you do it. But only because you let it.
👇 What do you think?
Is ChatGPT shaping how you write? Or just showing you how you’ve already started to conform?
References:
Neuman, S.B. (1995). Literacy in the Television Age. Praeger.
Yakura, H., Lopez-Lopez, E., Brinkmann, L., Serna, I., Gupta, P. and Rahwan, I. (2024). Empirical evidence of Large Language Model’s influence on human spoken communication. arXiv (Cornell University). doi:https://doi.org/10.48550/arxiv.2409.01754.
Liang, W., Zhang, Y., Wu, Z., Lepp, H., Ji, W., Zhao, X., Cao, H., Liu, S., He, S., Huang, Z., Yang, D., Potts, C., Manning, C.D. and Zou, J.Y. (2024). Mapping the Increasing Use of LLMs in Scientific Papers. arXiv (Cornell University). doi:https://doi.org/10.48550/arxiv.2404.01268.