← AI Insights
繁體中文 English 简体中文
Levi · LinkedIn · 2026-09-25

AI Use Boundaries for DSE, IELTS and CFA Prep

CFA has a clear stable boundary, IELTS has a specific scoring policy change, and the official DSE boundary is still pending a complete framework

DSEIELTSCFAAI Exam PrepExam Policy

The suitability of AI tools for exam preparation varies significantly by exam type and mode of use. Using AI effectively for exam prep depends on understanding each exam's current policy boundary, and which parts of preparation AI genuinely helps with versus where it carries a risk of misleading.

The DSE's Current Framework

The Hong Kong Examinations and Assessment Authority (HKEAA) has so far not published a standalone AI policy document. Practical guidance comes mainly from guidance the Digital Policy Office published in April 2025: students must obtain prior teacher approval before using generative AI in homework; the accuracy of AI-generated content must be confirmed by the teacher; AI must not directly generate submitted work.

The English Language School-based Assessment (SBA, 15% of the English Language total) already has specific guidance: using AI for brainstorming, planning and searching background information is appropriate use, while directly generating submitted content is inappropriate use. Classical Chinese translation and language comprehension in the Chinese Language subject are task types widely recognised in the industry as high-difficulty for AI, but there is no official accuracy assessment yet.

Reasonable AI applications for exam prep: generating practice questions (categorised by type and difficulty), explaining concepts, analysing question patterns in past papers. A misuse to avoid: using AI-generated model essays as the primary material for writing practice, since this trains recognition rather than production ability.

IELTS: The 2026 Template Crackdown

IELTS is invigilated at test centres, and AI tools cannot be used in the actual exam; the British Council and IDP currently have no clear official policy document on AI-assisted preparation.

However, a specific 2026 policy change directly affects preparation strategy: IELTS writing and speaking examiners have been trained specifically to identify memorised templates — AI-generated formulaic essays are now marked down in scoring. This means the strategy of using AI-generated essays as preparation material carries a real risk of backfiring after 2026.

Effective uses of AI in IELTS preparation: practising mock questions and answers for Speaking Part 1/2/3, getting writing structure suggestions and grammar checks, and having the assessment dimensions of the scoring criteria explained. What to avoid: using AI-generated model essays as a model for writing style to imitate, since examiners are now equipped to recognise this kind of output.

CFA: Curriculum Stability Confirmed for 2026

The CFA exam is taken on computer at test centres, and AI tools cannot be used during the exam. CFA Institute's adjustments to the 2026 curriculum are extremely limited: zero changes at Level 1; Level 2 removes only one LOS from the machine learning module (neural networks / deep learning, under 0.5% of the curriculum); zero changes at Level 3; and 2025 preparation materials can safely carry over to the 2026 exam.

CFA Institute's official position on using AI tools for out-of-exam self-study has not been made public, but using AI to explain concepts, generate practice questions and summarise reading material is a common preparation practice, with the boundary sitting at the exam room's strict rules.

Effective uses of AI in CFA preparation: explaining formula derivations and conceptual background, generating scenario-based practice questions, and summarising long textbook chapters to help manage review pacing. Worth noting: answers to AI-generated practice questions must be checked against the official curriculum material, and AI's familiarity with CFA-specific question formats varies with training data.

Summary

AI policy maturity differs across the three exams: CFA's curriculum is stable with a relatively clear boundary, IELTS has a specific scoring policy adjustment (template mark-down), and the DSE's official boundary is still pending a complete framework. The shared principle: AI is for practice and comprehension aid, not for generating content submitted directly or used during the exam itself.

For the overall education sector AI policy framework, see Hong Kong Education AI Policy: DSE, IELTS, CFA.

Levi is a Hong Kong-based independent AI engineer specialising in production LLM applications, RAG pipelines, and enterprise AI compliance architecture. Contact for a discussion of the topics covered here.

WhatsApp Free Initial Consultation → More enterprise case studies →

Or email: support@hksoka.com