So my school (if you can even call it that) has an external company provide the syllabus, provide and mark tests, exams and assignments (assignments are marked internally though). So they upload the unmarked tests and exams as soon as you write them and they are scanned in, and they upload digital copies of the question papers memos and the marked papers once results are out (although nothing is stopping me from scanning my own copy of the question paper) and this is all accessible from my portal. I’m wondering if you guys know any good AIs and tools that can mark my papers early instead of me having to wait with anxiety until exams are over and even longer till they actually get to marked (they tend to take a really long time). Are they any free Ais (or at least with a free trail) out there that mark tests and essays well, if I give it the question paper and the scan of my answer sheet. I’ve tried using chatgpt but there’s the upload limit and the general ai hallucinations that come with it. I also tried gemeni not that big of an upload limit but it tends to hallucinate stuff way more even in other applications. One thing is essays are way more important for me as they make up a lot of marks in most of my subjects, so an ai that’s good at that is a major positive.
Any help, ideas or suggestions would be much appreciated.
But you have no way of knowing what their grader is going to say is a correct answer.
Any AI solution is going to grade your work based on the data that was in its training set. There’s no guarantee the AI will be accurate or agree with the school’s grader.
I agree with you but at least having some sort of ball park number should be possible, plus if the context window is large enough I could just upload my textbooks for each exam (anywhere from 150 to 350 pages) so it can try make reference to that data.
I understand why you’d want an AI tool to estimate your marks early, especially when waiting for results can be stressful. However, I would be cautious about relying on AI for grading.
In higher education, exams are often not returned to students immediately, and in some cases they are not returned at all. Even when question papers and marking schemes are available, grading is rarely a completely objective process. For essays, open-ended questions, proofs, and partially correct solutions, different markers may award different amounts of credit based on their judgment of the student’s understanding.
Many institutions also use moderation processes to ensure consistency across markers, and in some cases grading can be influenced by broader assessment policies. An AI has no access to these considerations and can only make an educated guess based on the information you provide.
Because of this, an AI-generated mark may create either false confidence or unnecessary anxiety. I think AI is much more useful as a study tool—helping you identify missing points, weaknesses in an essay, or areas where marks could be lost—rather than trying to predict the exact grade you will receive.
Ultimately, the official result will depend on the human marker, the marking guidelines, and the institution’s assessment process, none of which an AI can fully replicate.
I work on AI systems for grading handwritten exams, so I can add a bit of perspective from the engineering side.
I agree that an AI can’t reliably predict the exact mark you’ll receive because every institution has different marking schemes, moderation policies, and expectations around partial credit.
That said, I do think AI can be useful as an estimation tool if you provide:
the question paper,
the marking scheme or answer key (if available),
and a good scan of your answers.
In my experience, the biggest source of error isn’t usually the model’s subject knowledge it’s understanding handwritten pages correctly. Students often answer questions out of order, continue answers on later pages, draw diagrams, or cross things out, which makes automatic evaluation surprisingly difficult.
I’d use AI to identify missing points or weak arguments rather than treating the final score as a prediction of your official grade. That tends to be much more reliable and useful.