Plenty of students now paste an answer into ChatGPT, Claude or Gemini and ask, "What mark would this get?" It is a sensible instinct, and chatbots are genuinely good study tools. But a mark is only as meaningful as the marking guideline behind it. This post explains what changes when the marker is given the official criteria for the exact question, and how to get much better results from a general chatbot if you want to keep using one.
Disclosure: this post is written by AusGrader. AusGrader also uses AI models, including models from OpenAI and Google. This is not a post about one AI being better than another. It is about the context the marker is given. If anything here is out of date, please reach out and we will correct it promptly.
1. The Short Answer
Chatbots are excellent tutors. They can explain a concept three different ways, quiz you, and turn a confusing paragraph into plain English. For that kind of work, keep using them.
Marking is a slightly different job. It works best when the marker has the question's marking guideline in front of it: the official criteria that say where each mark is awarded. AusGrader attaches that guideline to every question automatically, so the marker always starts from the same place.
The difference is context, not intelligence. Give any capable AI the question, the marks available and the official guideline, and it will mark far more usefully. The rest of this post shows how.
2. What Happens When You Paste an Answer Into a Chatbot
None of these are flaws in the chatbot. They are what you would expect from any marker who has not been handed the marking guideline.
- It marks against its own idea of a good answer. Without the official criteria, it has to guess what the examiners wanted. Its guess is often reasonable, but it may reward points the guideline does not, or miss a specific term the guideline requires.
- Marks can drift between attempts. With no fixed rubric to anchor to, asking the same question twice can produce a 3 out of 4 one time and a 4 out of 4 the next. That makes it hard to tell whether you improved or the marker changed.
- It does not keep your attempt history. A chat is a conversation, not a record. It will not show you which topics keep costing you marks across weeks of practice unless you track that yourself.
Each of these gets smaller the more context you provide, which is the idea behind section 4.
3. How AusGrader Marks
AusGrader uses AI too, but it does not ask the AI to work from nothing. For each answer, the grader is given:
- The question, and any stimulus (the graph, passage or data that goes with it).
- The official marking criteria for that question.
- The sample answer, used as a guide to what a full-mark response looks like, not as the only acceptable answer.
The grader is also set up to behave the way a careful marker would:
- Vague answers do not get marks. Your answer needs to say the thing the criterion asks for.
- Small differences are allowed. Rounding differences and keyboard notation (for example typing x^2 for x squared) do not cost you marks.
- The best valid method counts. Where several valid methods exist, the one that scores best is the one that counts.
- Command terms are respected. "Explain" and "evaluate" ask for different things, and the marking reflects that. Our command words guide covers what each one expects.
Handwriting. If you photograph a handwritten answer, AusGrader transcribes it first and shows you the transcription, so you can see exactly what the marker read. Crossed-out work is ignored, and any part it cannot read is marked as illegible rather than guessed.
Feedback and analytics. You get feedback per criterion, so you can see where marks were gained or lost. Your results then feed topic analytics, which show the topics that are costing you the most marks. For more on using marking to study, see how AI grading helps you study smarter.
It is still AI. AusGrader's marking is a strong guide, not an official result. Read the feedback critically, check it against the marking criteria shown on the question, and ask your teacher if a mark looks wrong.
Cost. Questions, sample answers and marking criteria are free. You get 10 free AI grades, then it is $100 one-time for 2026 ($50 in NSW). Self-marking, where you tick off the criteria yourself, is coming soon (no date yet).
4. How to Use ChatGPT Well for Marking
If you prefer to use a general chatbot, you can close most of the gap yourself. This works with ChatGPT, Claude, Gemini or any similar tool:
- Paste the exact question, including any stimulus or data, word for word.
- Tell it the marks available so it knows how much detail the question is worth.
- Paste the official marking guideline and the sample answer. Both are free on AusGrader (browse the VCE, HSC, QCE or WACE question banks) or on your board's website.
- Ask it to mark criterion by criterion and to quote the evidence from your answer for each mark it awards or withholds. Quoted evidence makes it much easier to spot a mark that was given too generously.
- Run it again to check consistency. If the mark changes between runs, treat the result as a range, not a score.
- Type out handwriting first. Check the text matches what you wrote before you ask for a mark, so a misread word does not cost you.
For a wider look at what else is out there, see our guide to AI study tools for Australian students, and how to mark your own practice exams if you want to learn the marking skill yourself.
5. At a Glance
| ChatGPT, Claude or Gemini | AusGrader | |
|---|---|---|
| Knows the official marking guideline | Only if you paste it | Attached to every question |
| Sample answer | Only if you paste it | Yes, used as a guide |
| Same rubric every time | Depends on your prompt | Fixed per question |
| Handwriting | Reads photos | Transcribes, shows you, then marks |
| Progress tracking | None by default | Topic analytics |
| Explaining concepts | Excellent | Feedback focused on the question |
| English essays | Yes, with your rubric | VCE, HSC and WACE English coming in 2027 |
| Cost | Free tier available | 10 free grades, then $100 ($50 NSW) |
6. Essays and English
This is where a general chatbot is the better option today. English for VCE, HSC and WACE arrives on AusGrader in 2027, so we do not mark English essays yet.
In the meantime, the method in section 4 works well for English. Paste your essay, your board's published assessment criteria and the prompt, and ask the chatbot to respond criterion by criterion with quoted evidence. Treat the result as feedback to think about, not a predicted score, and show your teacher the same essay when you can.
7. What AI Use Is Allowed
Marking your own practice answers is study, not submitted work. But assessment tasks that count towards your results have their own rules, and those rules are set by your board and your school. A few starting points:
- NSW (HSC): NESA says unapproved AI use in assessment breaches academic integrity, and any AI use must be acknowledged. See NESA's AI guidance.
- Victoria (VCE): VCAA updated its authentication rules for generative AI. See the VCAA notice to schools.
- Queensland (QCE): QCAA has AI guidance, and schools set their own rules on top of it. See QCAA's AI policy page.
- Western Australia (WACE): check with your school.
When in doubt, ask your teacher before using any AI tool on work that is assessed.
8. Coverage Note
AusGrader covers VCE, QCE and WACE across the sciences, maths, business, humanities and PE. For the HSC, today it is Biology, Chemistry and Physics only, and we are expanding to Maths, English, HSIE and PDHPE. English for VCE, HSC and WACE arrives in 2027. If your subject is not covered yet, the chatbot method in section 4 is a good option for now.
FAQs
Can ChatGPT mark my HSC or VCE answers accurately?
It can give useful feedback, especially if you paste in the question, the marks available, the official marking guideline and the sample answer. Without those, it marks against its own idea of a good answer, so the score can differ from what examiners would give and can vary between attempts. Treat any AI mark as a guide.
Does AusGrader use AI too?
Yes. AusGrader uses AI models, including models from OpenAI and Google. The difference is that the marker is automatically given the question, any stimulus, the official marking criteria and the sample answer for every attempt.
Is it allowed to use ChatGPT to mark practice work?
Practice marking is not submitted work, so it is generally fine. Rules do apply to assessment tasks that count towards your results: NESA, VCAA and QCAA each have AI guidance, and schools can add their own. Check with your school if you are unsure.
Can AusGrader mark English essays?
Not yet for VCE, HSC and WACE. English for those boards is coming in 2027. Until then, you can use a general chatbot with your board's published criteria and ask for criterion by criterion feedback.
Where do I get marking guidelines to paste into ChatGPT?
Your board publishes past exams and marking guidelines on its website, and AusGrader shows the official marking criteria and sample answer on every question for free.
Marking With the Guideline Built In
Try it on a question you have already attempted and compare the result with your chatbot.
- Official marking criteria and sample answers on every question, free.
- Feedback per criterion, so you see where marks were gained and lost.
- Handwriting transcribed and shown to you before it is marked.
- Topic analytics that show where your marks are going.
Free to start. 10 graded questions included, no credit card required.
The Bottom Line
ChatGPT, Claude and Gemini are great for learning and a decent marker when you give them the official guideline. If you do not want to paste that in every time, AusGrader does it for you, keeps the rubric the same on every attempt, and tracks your results by topic. Use whichever fits how you study, and keep reading the feedback critically.
If you are also weighing up a printed study guide, see AusGrader vs Cambridge Checkpoints. Pricing details are on the pricing page.