VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

How G3 G2 G1 Works | SLS Learning Assistant vs ChatGPT: Which Helps Students Learn Without Copying Homework?

Quick answer, updated 6 November 2026. Singapore secondary students studying G1, G2 or G3 subjects may encounter the Singapore Student Learning Space (SLS) Learning Assistant, a teacher-enabled AI learning tool, as well as general-purpose services such as ChatGPT. Neither tool should replace the student’s own reading, reasoning or assessed work. A useful AI study session ends with the student able to solve or explain a new question without the AI.

The key practical difference is context and control. MOE’s SLS Learning Assistant, known as SALiS when enabled, is set up within a school learning activity; teachers can select learning materials, set interaction limits and review discussions. A general-purpose AI chat such as ChatGPT can be used for many tasks but is not automatically aligned to your child’s school assignment, G-level or teacher’s marking expectations. Its responses also require checking. Access to any outside AI service must follow the school’s instructions, the service’s age rules and the family’s privacy decisions.

If a teacher has assigned a task in SLS, begin there. If the teacher says the work must be independent or AI use is restricted, follow that instruction. If an AI explanation is allowed, treat the tool as a source of questions and provisional explanations, not as the author of work submitted under the child’s name.

For general Full SBB background, read eduKate’s G1/G2/G3 Secondary Education guide. This article is about a different decision: how a student can use an increasingly capable AI tool without losing the ability to learn.

Find the answer you need

1. SLS Learning Assistant and ChatGPT do not perform exactly the same role

Singapore’s SLS AI programme includes learning and feedback features designed for use in the school learning environment. According to the October 2026 SLS guidance, teachers can enable the Learning Assistant for activities, set interaction limits and review students’ conversation logs. Students can use SALiS when it is enabled in an interactive component, after the relevant AI learning modules have been completed. Its availability is not a promise that every subject teacher has turned it on in every class.

ChatGPT is a general-purpose AI assistant. It can explain ideas, ask questions and support practice when used appropriately, including in learning-oriented interactions. However, a fluent response is not proof of correctness, syllabus alignment or school permission to submit it. A child should not assume the chatbot knows which exact G2 Science topic the teacher is assessing or which marking points the school expects.

QuestionSLS Learning Assistant when enabledGeneral AI chat such as ChatGPT
Who structures the activity?Teacher within the SLS learning setupUser’s questions and whatever materials are responsibly provided
Is it automatically a school assignment?May be included in a teacher-assigned SLS activityNo; external use needs separate school permission where relevant
Can responses be inaccurate?Yes; SLS explicitly warns that AI responses can be wrongYes; even confident, detailed explanations can be wrong
Is the tool automatically tied to a student’s exact G-level?Teacher sets the relevant activity and contextNo; the user must identify the syllabus and verify the response
Who can review the conversation?The SLS guidance says the relevant teacher can review student interactionsNot the school by default; service privacy and account settings apply
What remains the learner’s responsibility?Think, check, explain and complete work appropriatelyThink, check, explain and complete work appropriately

Do not confuse the SLS Learning Assistant with every other feature in SLS. MOE also describes an Adaptive Learning System, a Mathematics feedback assistant and tools offering annotations or short-answer feedback. The teacher may use a different feature for a different task. See MOE’s current overview of AI-enabled SLS features.

2. Why the difference matters for a parent

Suppose your child has a school Science assignment. An SLS tool is offered inside the activity, and the teacher asks students to use its questions to improve their explanation. That use is part of the assigned learning process.

Now suppose the child opens an unrelated AI chatbot, pastes the entire task and asks for a perfect answer to submit. Even if the answer happens to be correct, the process may bypass the skill the teacher is assessing. If the school has prohibited outside AI-generated submissions, it may also breach the assignment rules.

The relevant question is not “Which chatbot writes better?” It is “Which interaction makes the student’s own ability stronger while remaining within the assignment rules?” Technology should not conceal the answer to that question.

3. Does G1, G2 or G3 require a different AI?

No particular AI tool becomes educationally correct simply because a learner is taking G1, G2 or G3. These are curriculum and examination subject levels under Full Subject-Based Banding, mapped from the former Normal (Technical), Normal (Academic) and Express standards. They are not ratings of which device or chatbot a child deserves.

The content and demand of an explanation should match the actual syllabus. A learner taking G1 Mathematics may benefit from clear practical examples and accurate representations of the required principle. A learner taking G2 may need help connecting an expression to a word problem. A learner taking G3 may need help choosing among methods in an unfamiliar task. These are possible teaching needs, not stereotypes about ability. The real starting point comes from schoolwork.

Ask the system to use language appropriate to the student’s current understanding, but never tell it to invent a made-up “G3 formula” or treat its output as an official syllabus. The authority is the relevant SEAB SEC syllabus or the school’s assigned curriculum, not the chatbot’s confidence.

4. A six-step AI study routine that protects independence

Step 1: Start with the child’s own attempt

Before opening an AI tool, let the student read the assignment, underline the task and write a beginning. It may be imperfect. That first attempt shows what the learner actually knows. If the student cannot start, ask them to write a question about the part they do not understand.

Step 2: Ask for a hint, not an answer to submit

A useful request is: “Do not solve my school question. Ask me what I know and give one hint at a time so I can solve a similar example myself.” This makes the assistance easier to control. If the AI still gives the answer directly, stop and return to an independent attempt rather than copying it.

Step 3: Check the idea against a trusted source

Use the school textbook, teacher’s example, official syllabus or carefully selected learning material. A generative AI model may produce a plausible explanation while using the wrong definition, notation, method or source. If the facts cannot be checked, do not rely on them as settled information.

Step 4: Explain the principle without looking at the chat

Close or hide the conversation. Ask the student to explain the rule or solve a parallel example. If they cannot, the AI session may have produced a sense of understanding without the underlying skill.

Step 5: Change the question

Try one fresh problem with a new context, number or representation. Ask the child to identify what stayed the same. This is more informative than repeating the exact problem the AI already worked through.

Step 6: Respect the school’s submission rules

If AI assistance is permitted, disclose it in the way the teacher requires. If an assignment requires unaided original work, do not submit AI-generated paragraphs, code, diagrams or solutions as the student’s own. If the instructions are unclear, ask the teacher before proceeding.

A good AI interaction leaves the learner more independent, not more dependent. A child who can produce beautiful answers only with the chatbot open still needs a bridge back to their own thinking.

5. Mathematics: AI can explain a method without owning the solution

Consider a learner stuck on 2(x + 4) = 18. A general AI tool may provide the correct solution quickly. But if the student copies the steps, they may not have learned what the bracket means. Instead, use it to generate a simple parallel example and guide the student to represent two groups of the same quantity.

An effective study question would be:

“I can solve 2x = 18, but I do not understand 2(x + 4) = 18. Do not give my answer. Use a simpler example to help me understand brackets. Ask me to explain each step and then give me one different equation to solve independently.”

The student should learn that 2(x + 4) means 2x + 8, so the equation becomes 2x + 8 = 18. But the goal is not merely this answer. It is a reliable understanding of distribution that works again in 3(x + 2), 4(x − 1) and appropriate word problems.

Be particularly careful when an AI service confidently manipulates fractions, algebra, statistics or diagrams. Verify calculations independently. MOE itself warns that some AI-generated feedback is not suitable for mathematical computation and identifies separate tools designed for mathematical working. The presence of “AI feedback” is not evidence that every answer is mathematically sound.

6. English: a polished paragraph can conceal weak writing

Suppose a student is asked to write an argument about whether mobile phones should be restricted at school. They give a chatbot the prompt and receive five polished paragraphs. If they submit those paragraphs, a teacher may be unable to judge the student’s own ability to choose a position, organise reasons, use evidence and revise sentences. The student may also be unable to reproduce the argument independently under examination conditions.

A better approach is to have the student produce a short outline first, then ask the AI to challenge the structure without writing the essay. For example:

“Here is my outline. Do not rewrite it or add an essay for me. Ask me three questions about whether my reasons are relevant, where I need evidence and what a reasonable counterargument might be.”

Then the student writes the revised version independently, following school rules on permitted assistance. If they cannot explain why one paragraph belongs after another, more polished wording will not solve the underlying problem.

AI can be useful for discussing the meaning of a word in context, but dictionary definitions, usage and literary interpretations should be verified. A word lives inside a sentence, a situation, an intention, a culture and a history. Replacing that whole field with one auto-generated synonym is not a reliable route to mastery.

7. Science: ask the AI to expose reasoning, then verify the science

Imagine a child learning why a lamp stays dark in an incomplete circuit. They know the sentence “a complete circuit is needed for current to flow” but cannot apply it to a changed diagram. Ask for an analogy or a check question that reveals the missing reasoning. Then return to the textbook and an actual circuit diagram.

Useful request:

“I am learning electric circuits. Ask me which parts must be connected for the lamp to light. Give one simple diagram description, then change one connection and ask me to predict the result. Do not tell me the answer until I explain my reasoning.”

For laboratory work, chemistry or practical safety, AI advice should not replace school instructions or appropriate adult supervision. Some explanations may be oversimplified or wrong in ways that matter physically. Use the subject syllabus, teacher-provided materials and safety procedures as the deciding references.

8. AI can shorten a task without shortening the learning gap

A student may finish a written task in five minutes with AI assistance rather than forty minutes unaided. That is a real difference in task-completion time. It does not establish that the student learned more. The more important test is whether they can later explain or apply the same principle without support.

Use four checks after an AI-assisted session:

  • Independent retrieval: can the learner recall the core idea?
  • Explanation: can they say why a method works?
  • Transfer: can they apply it when a question changes?
  • Correction: can they identify an error in a plausible-looking answer?

This is an especially important distinction for a student taking mixed G-level subjects. A fluent chatbot can hide that the English assignment demanded inference while the student used only copying, or that the Mathematics syllabus demanded algebraic reasoning while the AI did all the manipulation.

For a wider explanation of that learning problem, see why students understand classwork but get stuck on unseen questions.

9. What is acceptable help, and what risks becoming misrepresentation?

There is no single rule that makes every AI activity acceptable in every school assignment. A teacher may permit AI brainstorming for one exercise, prohibit it in another and require acknowledgement in a third. A coursework or examination component can have stricter rules than a casual home revision session. Follow the school and assessment owner’s instructions for the actual task.

Outside an assessed task, asking for an explanation of a concept or a new practice question is different from presenting AI-generated work as personally authored. In assessed work, even asking for an outline or edits may require permission. The student’s integrity depends on understanding what assistance is authorised and representing their contribution accurately.

For detailed integrity and citation issues, the established eduKate guide is Academic Integrity and Original Work Across G1, G2 and G3. This page focuses on one practical test: after AI use, can the student still demonstrate their own work?

10. A three-question test for AI-generated facts

Ask:

  1. Where did the information come from? A model may name a book, website or syllabus without actually supplying a reliable source. Confirm the original source exists.
  2. Does the explanation fit this subject and year? A generic answer may draw on the wrong country’s syllabus, an older assessment system or a more advanced subject level.
  3. Can the reasoning be checked independently? Recalculate a Mathematics result, inspect quoted evidence in the actual passage, or verify a Science mechanism against approved learning materials.

For MOE subjects, distinguish the school’s teaching materials from the national examination syllabus. For 2027 graduating school candidates, SEAB lists subject-level syllabuses separately for G1, G2 and G3. A chatbot that invents a syllabus code or merges standards should not be treated as authoritative.

For any AI system, including school-provided tools, fluency is not proof. The SLS student guidance explicitly instructs learners to reflect on SALiS responses and verify them with other sources. That warning is valuable educational advice even outside SLS.

11. Why the correct workflow is not simply “AI on” or “AI off”

Different learning stages have different needs. When the student is first exploring an idea, carefully supervised explanations or analogies may help. When the student is checking a factual claim, trusted sources and teacher feedback matter. When the student is preparing for independent examination conditions, the AI should not be present during the actual independent attempt.

Consider a sequence:

Learning stagePossible useful AI roleWhat the student must still do
Encounter a new ideaAsk a clarifying question or offer an analogyCheck the real definition and syllabus scope
Practise with guidanceGive one hint or ask a Socratic questionComplete the reasoning
Correct an errorSuggest where to inspect a stepIdentify and fix the actual mistake
Test independenceNo AI during the independent attemptSolve, write or explain alone
Reflect afterwardProvide a new comparison question where permittedVerify and record what changed in understanding

The sequence should be adjusted to a child’s developmental stage, access needs, actual assignment instructions and home circumstances. It is not an official SLS feature description.

12. Age rules, privacy and safe use

ChatGPT is not intended for children under 13. OpenAI’s published guidance says that users aged 13 to 18 need parental permission, and that if an educational interaction involving a younger child is necessary, an adult must conduct the interaction. Families should check the current OpenAI age guidance and school policy rather than assuming that a service available on a phone is automatically appropriate for every student.

Do not upload a child’s identifying records, private health information, other students’ details, passwords, private teacher feedback or confidential school documents to an outside service without a clear, authorised reason and appropriate protections. A cropped or paraphrased question is often enough to ask for an explanation, and a made-up example may be safer.

OpenAI provides age-appropriate teen protections and optional parental controls on eligible accounts. Those controls are not a parent surveillance service; linking accounts does not let a parent read the teen’s private conversation history. See the ChatGPT for Teens information for current features and limitations.

By contrast, the MOE SLS Learning Assistant runs within the school’s learning environment. The official student guide explains that interactions may be reviewed by the teacher and that the tool is enabled for the relevant learning component. This is one reason school-directed use and personal AI use should not be treated as the same activity.

13. What teachers, parents and students each control

Teachers decide what an assignment is assessing, which forms of assistance are permitted, the learning materials to use and how to review progress. Parents can ask how AI is used at school, set sensible home boundaries, protect privacy and notice whether AI has become an avoidance mechanism. Students should retain the central work: reading, questioning, solving, correcting, explaining and acknowledging support honestly.

The best home conversation may be surprisingly brief. Ask: “What did the AI help you understand?” Then: “Show me one question you can now do without it.” If the child cannot demonstrate a new independent ability, the next step is teaching or practice, not a celebration of the chatbot’s polished answer.

14. Four warning signs of AI dependence

  • The child cannot start a familiar homework task until an AI chat is open.
  • The written answer is sophisticated, but the child cannot explain its main claim or the working steps.
  • The child accepts an invented citation, scientific fact or mathematical procedure because the response sounds confident.
  • The student repeatedly submits finished work without a visible independent first attempt when original work is required.

These signs are prompts for a calm learning conversation, not automatic proof of dishonesty. Ask what the assignment permitted and what the child actually did. Then rebuild one independent task at a time.

Frequently asked questions

Is SLS Learning Assistant the same as ChatGPT?

No. MOE’s Learning Assistant is a teacher-enabled SLS feature with school learning context, interaction controls and teacher review arrangements. ChatGPT is a separate general-purpose service, with its own account, age, privacy and product rules.

Can a student at G1 use AI to learn a G2 topic?

AI can help explore concepts, but it does not determine subject-level eligibility, teaching sequence or whether a move is suitable. Begin with the child’s actual syllabus and school guidance. A higher-level extension should not replace essential current foundations.

Can I ask AI to check my Mathematics answer?

For permitted home practice, yes, but verify the calculation and steps independently. AI can make convincing mathematical mistakes. Do not treat an AI check as an official mark or as a substitute for teacher feedback on assessed work.

Can ChatGPT write my English essay and then I edit it?

Whether such assistance is authorised depends on the assignment. Even where permitted, that workflow risks letting the AI do the main writing work. A safer learning approach is to create your own outline and draft first and request questions or limited feedback within the teacher’s rules.

Does the school automatically see my ChatGPT conversations?

A separate personal ChatGPT account does not become an SLS conversation merely because the student attends a school. However, do not assume privacy when using school-managed devices or accounts; follow the school’s device rules. SLS Learning Assistant interactions have their own teacher-review arrangements.

What if the AI disagrees with the teacher?

Ask for the relevant source and compare it against the school syllabus or trusted reference. AI responses can be wrong or out of context. Raise an unresolved academic question with the teacher rather than treating AI confidence as a verdict.

Will AI replace tuition for my child?

Neither AI nor tuition is automatically required. Use school resources and diagnose the actual learning problem first. AI may provide a helpful practice interaction, but a persistent misconception, access need or workload difficulty may require teacher guidance or targeted human support. The relevant measure is improving the child’s independent learning.

Further reading and official resources

Source check: 6 November 2026. SLS descriptions reflect its official October 2026 student and teacher guidance, including AI error warnings, teacher-enabled interactions, limits and conversation review. OpenAI age requirements were checked against its published Help Center. MOE’s broader AI literacy direction is described in its 2026 announcements; UNESCO provides independent context on responsible AI use. The study sequences and sample prompts in this article are eduKate teaching suggestions, not official MOE or SEAB policy. AI features, school instructions and product settings can change; verify before use.