IB Geography EE Grader
AI grading for the IB Geography Extended Essay: predicted grade band and per-paragraph feedback. Geography EEs reward a spatially focused question answered...
Geography EEs reward a spatially focused question answered with evidence: fieldwork or data, maps used as analysis, and conclusions limited to the area studied.
What we grade for
- Spatial focus of the question
- Quality of primary and secondary data
- Maps and graphs used in the analysis
- Evaluation of the method
- Geographic terminology
Common pitfalls
- Question not spatial or not local enough to investigate
- Maps and graphs included but never discussed
- Secondary data quoted without evaluation
How the Geography EE is marked
- Framework for the essay (max 6) - Focused RQ, suitable research methods, a structure that carries the research
- Knowledge & understanding (max 6) - Relevant sources, accurate terminology, concepts explained and used
- Analysis and line of argument (max 6) - Analysis rather than description, one clear argument from RQ to conclusion
- Discussion and evaluation (max 8) - Balanced, evidenced discussion; strengths and limitations of your own essay - heaviest weight
- Reflection (max 4) - Marked from your reflective statement (max 500 words) - growth and transfer of learning
Frequently asked questions
What is the MyRevisionAgent Extended Essay grader?
An AI-powered grader that scores your IB Extended Essay against all five criteria first examined in May 2027 - Framework for the essay, Knowledge and understanding, Analysis and line of argument, Discussion and evaluation, Reflection - out of 30, or on the previous 34-mark criteria if you sit in November 2026. You get a predicted band, criterion-by-criterion justification, per-paragraph feedback, and a confidence range.
Which EE subjects does it support?
21 subjects: Biology, Business Management, Chemistry, Classical Languages, Computer Science, Design Technology, Digital Society, Economics, English A, English B, Environmental Systems and Societies, Geography, Global Politics, History, Mathematics, Philosophy, Physics, Psychology, Social and Cultural Anthropology, Sports, Exercise and Health Science, and World Religions. 10 of them have examiner-marked training essays; the rest are graded by the same model with a wider error range and a note that we have fewer training essays for that subject.
How accurate is the AI grading compared to a real IB examiner?
Our model is trained on 593 real, IB-awarded Extended Essays. On held-out essays it lands within one grade band of the examiner 96% of the time across the full four-seed cross-validation (47% exact-band match), with a mean absolute error of about 3.6 marks out of 34 - and every prediction ships with an explicit ±error range so you can see the model's confidence. Frontier LLMs used as a generic prompt aren't calibrated to the IB mark scale: on the same 116 held-out essays used for the head-to-head, ChatGPT (GPT-5.5) landed within one band 79% of the time and Claude Opus 4.8 93%, against our 97%. These figures are measured on essays marked under the criteria in use until November 2026: the 30-mark criteria are first examined in May 2027, so there are no examiner marks on them to test against yet.
Is it free? Do I get a free Extended Essay grade?
Yes - we add 2 free grading tokens to every new account on signup, usable on any grader including the EE. After that a grade is £9.99 pay-as-you-go (cheaper in packs), or £25.99 a month for five grades on the plan, which also gets premium feedback. Resources, guides and the in-app rubric are free to browse.
Is using an AI grader allowed by the IB? Will it flag me for academic misconduct?
Using an AI tool to get feedback on your own draft is fine and is treated the same as a teacher or tutor giving you feedback - the IB's academic-integrity policy is about not submitting AI-generated work as your own. We never write your essay for you. We grade your essay and explain why, so you can revise it yourself. Always disclose AI tool use as instructed by your school.
Does the grader assess my research question?
Yes. Research-question alignment is one of the seven subject-aware features the model extracts. It checks whether your RQ is appropriately scoped, focused, and analytically framed for the subject - and feedback flags exactly where the RQ is letting Criterion A down.
Do I get per-paragraph feedback on my essay?
Yes. Every flagged clause is highlighted on your essay in the colour of the skill it concerns - focus & RQ, analysis & argument, evidence & method, knowledge & terms, structure & clarity - with a severity label and a specific revision note. Clauses that already earn their marks get a green check. Click any highlight to jump to its note.
What is the maximum essay length the grader accepts?
The grader needs at least 1,500 words and accepts up to 80,000 characters - a full 4,000-word EE plus your bibliography and appendices fits comfortably. Anything outside that range is stopped before grading, with a clear message, so you don't spend a token on it.
How long does grading my EE take?
Grading takes 4-6 minutes end-to-end, including reading your PDF: our layout model reads data tables as tables, and an AI vision pass transcribes equations and adds a short description of each graph or figure. Progress streams live through every stage, so you always see exactly where your grade is.
How is this different from just asking ChatGPT or Claude to grade my essay?
A generic LLM hasn't been trained on the IB's actual mark scheme or thousands of real IB-awarded essays. It will write fluent prose about your essay but its predicted mark is essentially a guess. Our grader pairs a frontier LLM (for the close reading and feature extraction) with a classical ML model - gradient-boosted trees and ridge regression - trained on real IB-awarded EEs, cross-checked by a statistical validator that flags when the two disagree. Every prediction ships with an explicit uncertainty range and a confidence score, and we flag it on your grade whenever the validator diverges from the base model.
Is my Extended Essay used to train your models?
No. Submissions are never used for training. They are processed to produce your grade and stored only against your own account history so you can revisit them.