If AI Can Write the Essay, What Should We Be Assessing?

Recent studies have argued that AI can outperform the average human, and even highly qualified, top-quartile-performing humans, on basic academic tasks such as item response, but also on more subtle analytical tasks such as essay writing. Of course, what makes an essay truly brilliant—that flair of genius and probing insight—is difficult to define objectively. Despite efforts to standardize marking through assessment moderation, examiners inevitably differ in how they evaluate qualities such as structure, analysis, and originality. Yet even in these more subjective domains, AI often produces writing of remarkably high quality.

Yet this distinction is important. Producing a high-quality piece of writing is not necessarily the same as developing the intellectual habits required to create it. An essay is valuable not simply because it results in polished prose, but because of the reasoning, synthesis, reflection, and judgment that emerge through the process of writing it.

So the question is: if AI can write the essay, what should we be assessing? Should we still be giving essays to students?

To me, the answer is yes, we should keep giving essays to students in class, because the essay is a powerful learning experience that refines discourse, structures thought, and coaches long-term memory, rhetoric, and analysis. To be brought to a high level of essay writing is to exercise higher-order thinking and important cognitive skills. It also trains the mind to engage in serious discussion in a substantiated manner, which is important if we want society to be critically alert and for its citizens to be able to engage in serious debate, circumspection, doubt, consideration, exploration, and research.

Writing also forces students to organize ideas, evaluate evidence, and defend positions in a way that few other forms of assessment require. These are not simply academic exercises; they are habits of mind that underpin responsible citizenship, effective leadership, and lifelong learning.

The fact that AI can do better than the average human is, in many ways, neither here nor there. Do we stop playing chess once Deep Blue beats Garry Kasparov? Do we stop going to the gym because a hydraulic pump can lift more weight than a human? Clearly not, because the goal is human flourishing, not competing with machines.

The deeper question is how to integrate AI into our teaching and learning, knowing when to use it, when not to use it, and when to leave it aside. If we want to ensure that students are writing essays with no assistance, then those essays should be done in real time and checked for AI input. If we wish to see how students can integrate AI into their work, we should allow them to do that and write it into the task.

Different assessments may rightly have different purposes. Some should measure a student’s independent reasoning without technological assistance, while others may deliberately evaluate how effectively students can work alongside AI. The important point is that assessment should reflect the intended learning outcomes rather than the technology available to students.

I asked AI to read what I’ve just written and compose a two-sentence conclusion. Here it is, with a few human tweaks from me:

Ultimately, the educational value of an essay lies not simply in the finished piece of writing, but in the reasoning, synthesis, reflection, and judgment that emerge through the act of composing it. By intentionally balancing unassisted, in-class writing with thoughtful AI integration, educators are able to evaluate both foundational human thought and a student’s ability to leverage modern tools with discernment and responsibility.