Responsible AI Use
This page outlines how our AI Assistants function, what data they rely on, known limitations, and a changelog of updates.
At Exam Insights, we’re committed to the responsible use of AI in education. Our AI Assistants are built to support student learning and teacher impact — never to replace the role of educators or make automated decisions about student performance.
Our AI Assistants
Exam Insights includes a suite of AI Assistants, each designed with a clear and limited purpose: guiding students through complex exam questions, providing feedback on responses, breaking down syllabus content, and clarifying formulas. These assistants are powered by large language models (LLMs) and operate within strict system prompts and parameters to ensure responses are aligned with official syllabus materials.
AI responses are generated in real time and do not persist across sessions. Each interaction is independent, ensuring privacy and safety are preserved.
We do not use AI to make determinations about student ability or progression. All AI-generated content is intended to guide understanding — not to replace teacher judgement.
To explore the features and intended uses of each assistant, visit the AI Assistants page.
Data Sources
Our AI Assistants are built on large language models (LLMs) that have been pre-trained on a broad dataset, which may include publicly available internet text. However, within the Exam Insights platform, we apply strict prompting, domain constraints, and contextual filtering to ensure all AI responses are grounded in verified, syllabus-aligned material.
To guide responses and maintain curriculum relevance, our AI is instructed using:
- QCAA subject syllabuses
- QCAA official marking schemes
- QCAA formula sheets
- Internally curated metadata (topics, subtopics, cognitive verbs, and learning objectives)
- Structured classification of past exam questions within our platform
We do not allow AI Assistants to answer questions outside the scope of supported syllabuses, and we actively monitor and test responses to ensure accuracy, relevance, and educational appropriateness.
Known Limitations
While our AI Assistants are carefully designed and regularly reviewed, there are current limitations you should be aware of:
- AI usage tracking not yet available: While our AI Assistants are actively used across the platform, we do not currently track or report on individual or school-level AI usage. A real-time classroom dashboard with AI-specific engagement insights is in development and will allow schools to monitor how students are interacting with AI features once released.
- Syllabus-bound responses only: AI will not answer questions outside the syllabus or provide speculative, non-educational responses.
- No session memory: Conversations and grading feedback are not saved or retrievable once the session ends. Users should copy responses they wish to keep.
- Formula ambiguity handling: When formulas contain multiple unknowns, Formula AI may request clarification before proceeding.
- Usage limits on free accounts: Free users have a capped number of queries per week. Unlimited access is available via our Pro plan.
Changelog
We maintain a changelog of significant updates to ensure transparency and traceability in how our AI features evolve over time.
Enhancements to Grade with AI and Formula AI
Grade with AI has been improved to deliver clearer, more precise feedback — particularly when partial marks are awarded or missed. In addition, Formula AI now prompts students to specify which variable they’re solving for when a formula involves multiple unknowns, ensuring greater accuracy and relevance in responses.
Expanded AI Capabilities Across the Platform
We introduced Syllabus AI, a new assistant designed to help students explore and understand syllabus content at the unit, topic, and learning objective level. At the same time, AI Tutor was enhanced to support follow-up queries without resetting the session, and Grade with AI was launched in beta, providing real-time, syllabus-aligned marking for extended response questions.
Our Commitment to Safe and Transparent AI
We design AI to support human learning, not replace it. All Exam Insights assistants are carefully scoped, syllabus-aligned, and rigorously tested to ensure they are helpful, accurate, and appropriate for educational use.
We uphold best practices in:
- Student safety and data privacy
- Clear boundaries and explainable AI behaviour
- Transparent communication about AI use and limitations
If you have any further questions about how your data is used, or how our tools align with your school’s policies, please contact us here.