Why ChatGPT can't grade your MBA essay

Every applicant we talk to has already pasted their essay into ChatGPT. They get back enthusiastic, well-structured feedback. The essay feels improved. Then it gets rejected.

Here is what happened. Generic large language models grade your essay against generic essay rules — grammar, structure, vivid language, a clear thesis. MBA admissions does not score against generic essay rules. It scores against a specific rubric that is different at every school. Three reasons ChatGPT misses.

1. No school-specific rubric

HBS scores for leadership impact. Stanford scores for authentic values. Wharton scores for collaboration and quantitative confidence. Booth scores for intellectual rigor. The same essay can earn an 8 at one school and a 4 at another, and the feedback should reflect that.

A generic model treats all five schools the same. It will tell you the essay is "well-structured and compelling" regardless of which program you are targeting. That is not feedback. That is a compliment.

1. No grounding in the failure modes that actually get rejected

HBS readers reject essays that read as resume narration. Stanford readers reject essays that name a value without revealing one. Wharton readers reject essays that claim collaboration without showing the group. These are specific, named failure modes — and they are the patterns that decide most files.

A generic model does not know these. It will praise the resume-narration paragraph for being "clear and well-organized." That praise is technically true and admissions-wise fatal.

1. Trained to flatter, not to grade

General-purpose chat models are tuned to be helpful, agreeable, and encouraging. That tuning is the opposite of what you need from a grader. You need someone who will tell you the second paragraph reads as ego, the closing line is a cliche, and the scene in the third paragraph never actually happens.

What we built instead

MBA Essay Critic uses a rubric per school. The rubric is loaded into context at grade time so the model is graded against — not just informed by — the actual evaluation criteria the admissions board uses. The output is a score with a verdict, paragraph-level criticism, and rewrites grounded in the rubric, not in generic essay craft.

It is not perfect. A human strategist still wins for early-stage positioning, when you do not yet know what to write about. But once you have a draft, rubric-aligned scoring is the fastest and most consistent way to iterate. And it costs $79 instead of $500 an hour.