Why Academic Writing Is So Bad, and What It Has to Do With AI

If you've ever read an academic paper and wondered why the authors seemed to be actively hiding what they meant, you're not imagining things. Academic writing developed specific conventions over the twentieth century that prioritize the appearance of rigor over actual clarity. Those conventions became stylistic defaults. Now they're so embedded that people who write them have stopped noticing.

Where the style came from

Academic writing became formal and impersonal largely because it was trying to seem scientific. The passive voice implies objectivity — not "I concluded" but "it was concluded." Hedging implies caution — "this may suggest" rather than "this shows." Long sentences with subordinate clauses signal complexity and depth, whether or not the idea is actually complex.

These conventions made sense as overcorrections against early scientific writing that was too speculative and personal. They calcified into a style that now exists independent of their original justification.

What it costs

Steven Pinker, who studies language and writes about it clearly, has argued that academics write in a deliberately complex style partly to signal membership in a community of experts. The prose is a badge. Clear writing is sometimes seen as superficial, while deliberately obscure writing seems profound.

The practical cost is that ideas that could change how people think about something stay inside the discipline. The research is real. The writing keeps it from traveling.

The AI connection

Language models were trained on enormous amounts of text. A significant portion of that text is academic. Models learned academic conventions the same way they learned everything else — from exposure.

This is part of why AI writing feels hedgy and impersonal even when asked to be clear. It defaulted to the style that appeared most often in its training data. "It is important to note that" and "this study suggests" and "a number of scholars have argued" are all academic patterns that bled into general AI output.

What clear academic writing looks like

It exists. Oliver Sacks wrote about neurology as a humanist. Richard Feynman explained physics in language that made you feel like you'd thought of the idea yourself. Temple Grandin writes about autism and animal cognition from the inside of her own experience.

The through-line is that they had a reader in mind — a specific, curious person who didn't already know the material — and they wrote for that person rather than for peer reviewers. That choice changes everything about the prose.

Most academic writing doesn't make that choice. AI writing didn't learn to make it either. The fix, for both, is the same: decide who you're talking to, then actually talk to them.