What makes human writing worth reading?

Suppose I wanted a short explanation of generative artificial intelligence.

I asked DeepSeek to write one for me. Here is what it produced: Generative AI uses complex algorithms to solve problems, draft essays and generate images. While it can improve efficiency, it may also reproduce bias or produce inaccurate information. Therefore, society should balance technological innovation with careful consideration of its risks.

I then asked a human writer to revise the paragraph. Here is the revised version: Generative AI can write poems, draw images and help solve scientific problems. But because it learns from human-made data, it can also carry human bias, like barely visible cracks in fine china. If we trust it blindly, the harm may not stay on the screen. It may affect how essays are graded, how people are judged or whose voices are taken seriously.

Take a minute to analyze how the two versions differ.

The first paragraph is not wrong. In fact, it is clear, balanced and fluent. But it also feels familiar. It could have been written for almost any audience, in almost any context. Its ideas are sensible, but they do not reveal much about the writer. We do not know what the author cares about or what they want us to notice.

The second paragraph does something different. It still discusses the strengths and risks of AI, but it makes sharper choices. It selects concrete examples. It uses a metaphor — "barely visible cracks in fine china" — to make the idea of hidden bias easier to grasp and remember. Most importantly, it shows judgment. The writer is not simply repeating a general view of AI, but guiding readers toward a particular concern: that convenient tools can also carry real-world harms.

This is where human writing still matters. While AI can help generate ideas or give feedback on your writing, good writing requires purpose, context, voice, emotion and creativity. A strong writer does not simply arrange fluent sentences — they make choices.

First, write with purpose. What do you want to accomplish? What do you want the readers to know or do? A student writing about AI, for example, may not only want to define the technology, but also warn readers against trusting it blindly. That purpose shapes what the writer includes and what they leave out.

Good writing also depends on context. What does your audience know? What are their priorities? A general paragraph about AI may sound correct, but it may not speak to a reader's real concerns. By choosing examples that connect AI to essay grades, scientific discovery or social justice, the author gives readers a reason to pay attention.

Then comes voice. Find your voice. What do you care about? This is something AI cannot easily provide. It can imitate style, but it does not have personal concern, memory or lived experience. The human writer decides why the subject deserves attention.

Good writing should also elicit emotion. What do you want your audience to feel? The image of hidden bias as "barely visible cracks in fine china" works because it turns an abstract problem into something fragile and visible. It makes readers sense a danger that might otherwise go unnoticed.

Finally, be creative. Try expressing the same idea in different ways. Use metaphors. Explore different settings. Creativity is not only about beautiful language. It is about finding a way of saying something that helps readers see it differently.

Written by Ming Ming Chiu, chair professor of analytics and diversity and director of the Analytics/Assessment Research Center, The Education University of Hong Kong.

Generative artificial intelligence tools such as ChatGPT and DeepSeek are now widely used by students to brainstorm ideas, polish language and complete writing tasks. These tools can quickly produce coherent and grammatically accurate text. Yet their writing often sounds neutral, predictable and detached.

As AI-generated essays become smoother, a new question has emerged: what still makes human writing worth reading?