Singapore’s Ministry of Education has developed an AI-assisted feedback tool called Markly. By September 2026, Geography teachers in Secondary schools and junior colleges could use it to review students’ extended written responses, while selected Secondary schools were piloting its use for English, Social Studies and History. MOE told The Straits Times that it would assess teachers’ and schools’ feedback before deciding the wider rollout. No publicly verified schedule says every Punggol school must use Markly in 2027.

For parents, the useful question is not whether a machine can mark an essay quickly. It is whether students receive feedback they can understand and act upon, then improve their next independent response. The announced approach keeps teachers responsible: they review, edit or reject AI suggestions before sharing anything with students. Markly supports the teacher’s judgement rather than replacing it.

That distinction is especially important for Humanities. A Geography explanation is not merely a tidy paragraph. It needs a valid relationship between causes and effects. A History answer needs sound interpretation of evidence. Social Studies needs reasoning from sources and claims. English writing needs purpose, audience and language control. This guide explains how a Punggol student can use an emerging feedback workflow without mistaking fluent automated comments for proof of learning.

What We Know About Markly in October 2026

In a report published on 5 September 2026, The Straits Times quoted MOE describing Markly as a tool developed to help teachers provide formative feedback on essays and extended written assignments. It was conceived in 2025, and its Geography use was available to teachers in Secondary schools and junior colleges. MOE said English, Social Studies and History applications were being piloted in some Secondary schools.

This information supports a particular status description. Geography is the established subject use mentioned in that report. Other named subjects were under trial. It does not support a claim that every subject, school or level already has full access. An article written in October 2026 should not invent a 2027 nationwide start date or an individual school’s participation.

The public example described in the report came from Canberra Secondary School, not a Punggol campus. It showed what iterative writing practice could look like when teacher oversight and AI feedback work together. Punggol families can learn from the teaching method without claiming that local schools have adopted the same system.

First Understand Formative Assessment

Formative assessment is feedback given during learning so the student can improve before the work is considered finished. Its purpose is different from a final grade. A teacher may use a draft, a response to a document, a short explanation or a classroom question to discover what the learner misunderstands.

Good formative feedback has to identify a change the student can make. ‘More detail needed’ may be accurate but vague. ‘Explain how the increase in impermeable surfaces leads to faster surface runoff after heavy rain’ gives a Geography learner a particular relationship to repair.

The difference matters in a classroom with limited time. Teachers have many scripts to read. A tool that makes routine annotation faster may allow more cycles of drafting and discussion. But if students simply copy suggested sentences, speed has improved without comprehension. The key measure is what the student can explain independently afterwards.

What Teachers Remain Responsible For

MOE’s reported position is that teachers review, edit or reject Markly’s suggestions before sharing feedback and remain responsible for the comments students receive. This matters for accuracy, context, fairness and the different needs of learners.

A Geography answer may use a correct but less common example. A History response may reach a defensible interpretation through a path unlike the model answer. A Social Studies learner may present a nuanced counterargument. A teacher can recognise when an unusual answer is supported by evidence rather than misclassify it as wrong.

There is also a personal dimension. A teacher knows which explanations the class has encountered, what support an individual student has already received and whether a comment will clarify or overwhelm. The machine is not the final authority on what to teach next.

Markly and AIMS Are Different Programmes

Families may also have heard of MOE’s AI Essay Marking System (AIMS), announced for Secondary Chinese on the Student Learning Space from September 2026. AIMS was described as supporting routine checks of grammar, spelling and sentence structure while teachers retain higher-order judgement.

Markly, as reported in September, supports formative feedback on Geography and selected pilots in English, History and Social Studies. Its task and pilot status should not be merged with those of AIMS. They share a principle of teacher review but have different publicly described subject scopes and reported implementation.

The distinction prevents confusion. A parent who sees an article saying ‘AI marking is available in schools’ should ask which named tool, for which subject, in which year and under what teacher controls. Without those questions, a valid statement about one programme becomes a false claim about every classroom.

What a Good Geography Explanation Must Contain

Take a familiar urban issue: why paved surfaces affect rainwater runoff. A weak response says, ‘There are buildings and roads, so there is flooding.’ The words are related to the topic, but the mechanism is missing.

A stronger response identifies that concrete and asphalt often reduce infiltration compared with permeable ground. More water can move over the surface towards drains during rainfall; if drainage capacity is exceeded under the conditions described, local flooding may result. The learner should explain these conditional relationships, not claim that every heavy rain must cause a flood.

An AI-assisted tool could potentially flag a missing explanation, while the teacher checks whether the causal chain matches the geography taught. The student then rewrites the relationship using a different scenario, such as comparing a paved courtyard with a planted area. If the new explanation is coherent, learning has transferred.

Worked Scenario: Describing Without Explaining

Imagine a Secondary 2 student studying the urban heat island effect. Their answer contains four observations: many buildings, little shade, busy roads and warm surfaces. It lacks the process linking those observations to temperature differences.

A useful diagnostic says: the student knows relevant features but is not expressing causation. Ask the learner to identify one mechanism, such as heat absorption and release by built surfaces, then connect it to the pattern of temperatures being discussed. A separate mechanism might concern reduced evaporative cooling where vegetation is sparse.

The learner first explains one relationship in their own words. Next they apply the idea to a different street or land-use comparison. This is stronger than adding a decorative phrase that makes the answer longer without making it scientifically sound.

The example is hypothetical. It illustrates the level of reasoning a teacher should look for, not a claim that Markly has a particular urban-heat-island scoring template.

Worked Scenario: A History Source That Disagrees with the Student

A Secondary 3 student reads two historical sources with different perspectives. Their response chooses the source they prefer and labels the other one unreliable merely because its conclusion differs. That is a reasoning problem, not a punctuation problem.

The teacher should ask who produced each source, when, for what purpose, with what audience and what corroborating evidence is available. A source may be limited in scope or partisan and still contain useful information about the period or the perspective it represents.

Feedback must help the learner distinguish the claims made by the source, the evidence supporting those claims and the student’s own evaluation. The student can then test the approach on a new pair of sources. A revised answer is meaningful when it demonstrates evidence-based judgement, not when it reproduces a highly polished evaluation written by another system.

Worked Scenario: Social Studies Evidence Without an Inference

A Social Studies learner quotes a sentence from a source and then says, ‘Therefore the government is good.’ The quoted text may be relevant, but the conclusion is too broad and is not justified by that single piece of evidence.

A better response explains what the source actually supports, its context and the limits of the inference. The student may acknowledge a particular policy benefit while recognising that the source does not prove every policy is effective.

The teacher’s feedback might therefore focus on inference precision rather than simply on a missing quote. The student practises with a fresh source and must select one claim that the evidence genuinely warrants. If they begin to avoid sweeping conclusions independently, the correction has produced an important analytical habit.

Worked Scenario: English Situational Writing and Audience

A Secondary student writes an email to a community organiser. The grammar is mostly correct, but the tone is too informal and the purpose arrives in the last line. A feedback tool might notice sentence errors while missing the most consequential communicative weakness.

The teacher asks: who is receiving the email, why are they reading it, what action is requested and which details are needed to make that action possible? The student moves the request near the opening, supplies relevant context and chooses an appropriately courteous tone.

On the next independent task, the student drafts a different message for a different audience. Success occurs when they adjust purpose and register without a correction being supplied. That is a valuable improvement in writing competence, not merely a cleaner first email.

The Reported Classroom Example from Canberra Secondary

The September report described a teacher at Canberra Secondary using Markly in an English situational-writing activity. The teacher provided several previously marked scripts so the system could learn the feedback style he used, and students completed multiple drafts while responding to comments.

This is a reported example of one teacher’s practice, not a universal requirement that every Markly teacher must upload the same number of samples or run six revisions. The broader educational point is that the teacher created a cycle of feedback and rewriting that would otherwise require substantial marking time.

The report also described the teacher choosing not to use Markly’s grading capability for that activity, preferring students to focus on improvement and receiving the final grade from him. Parents should not generalise that choice into a national rule; it is an example of how an educator may separate formative learning from summative grading.

Why Faster Feedback Is Useful Only Under Certain Conditions

Students often benefit from corrections while they still remember what they were trying to communicate. A comment returned weeks later can be hard to connect to the original thought process. Tools that shorten the interval between attempt and explanation may therefore create a useful opportunity.

But faster is not automatically better. An immediate stream of twenty suggestions may overwhelm a learner who only understands two. A correction that changes the author’s meaning can be actively harmful. And revising too quickly may encourage superficial editing instead of reflection.

Good feedback identifies a manageable priority, explains it clearly and gives the student space to try independently. A teacher decides when the learner needs another prompt and when a full new task would provide more valid evidence of improvement.

A Three-Step Diagnostic for Parents

First, inspect the original response. Was the central task understood? A student who answered the wrong question needs a different intervention from a student who understood the question but used weak evidence.

Second, locate the first break in reasoning. Did the explanation omit a mechanism, misuse a source, introduce an unsupported inference or present relevant facts in an incoherent order? Naming the break reduces random rewriting.

Third, have the learner repair that break and attempt a parallel question without opening the corrected answer. This confirms whether the correction has become a transferable skill. Keep the routine short enough to fit around real school responsibilities.

The parent does not need to be a subject specialist to ask the student to explain the correction. For technical correctness, rely on the relevant teacher and syllabus resources rather than guessing.

A Transfer Test for Geography

Suppose the learner has corrected an explanation of urban flood risk. Do not ask them to copy the paragraph on flooding again. Give a different situation involving soil, vegetation or surface characteristics and ask for a causal chain.

The child should identify the starting condition, describe the relevant process and state the outcome while noting limitations. For example, more vegetation can influence infiltration and interception, but actual outcomes also depend on rainfall intensity, soil conditions and drainage.

The new case checks understanding more fairly than repeated recall of a memorised sentence. A student who applies the mechanism appropriately can cope when an exam question changes its setting, diagram or wording.

A Transfer Test for Humanities Sources

After a student has revised a History or Social Studies source answer, select a second unfamiliar extract with a different perspective. Ask the learner to identify the claim, supporting detail, possible purpose and one reason to be careful about over-interpreting it.

Do not reveal a model answer immediately. The learner needs an opportunity to commit to an interpretation, test it against the source and revise if the evidence does not support it.

A short independent attempt followed by thoughtful discussion is more informative than an essay whose final form has been produced through repeated external prompts. The point of feedback is to build a reasoning habit that survives new source material.

A Transfer Test for English Writing

The learner has corrected a paragraph that introduced examples but failed to explain their relevance. Now ask them to write about a different issue and link one concrete example to a defensible point. They should be able to explain why the example supports the argument.

Alternatively, change the audience of a situational writing task. A request to a classmate may use different language from one sent to a school leader, even though the underlying information is similar.

These tasks reveal whether the student has learned a communicative principle rather than a single polished phrase. The best final response belongs to the learner.

Student Confidence and the Language of Corrections

Some students interpret every annotation as evidence they are poor at a subject. Others read feedback as instructions they must obey without understanding. Both habits can limit growth.

A teacher’s wording matters. ‘This paragraph has no point’ is less useful than ‘Your example is relevant, but the link to your central point is missing.’ The second statement identifies what is already working and specifies the next repair.

Students can learn to turn a comment into a question: ‘Which sentence should I add to explain how this happened?’ or ‘What evidence would justify this inference?’ It is reasonable to request clarification when an automated suggestion is too abstract or contains unfamiliar terminology.

How to Build a Small Feedback-to-Learning Log

Use four columns: task and date, first unstable step, the repair principle, and performance on a new attempt. A short record is sufficient. For a Geography task, the repair principle may be ‘state the process between cause and result’. For Social Studies, it may be ‘do not infer more than the source supports’.

Review the log after several tasks. If the same issue recurs, dedicate a short practice session to it. If a formerly common problem disappears on unfamiliar questions, retire that target and move on.

The log must not become a punitive scorecard. Its function is to make learning needs visible and to help the student notice their own improvement. Progress is not measured by how many correction entries accumulate.

When a Student Should Challenge a Comment

AI assistance can produce a suggestion that is plausible but inaccurate. The learner should compare it with the source, the question and classroom teaching. If it changes the meaning or assumes an unsupported fact, the student can explain their concern to the teacher.

This does not mean arguing with every correction. The proper habit is evidence-based checking. Ask, ‘Does this proposed wording still answer the question?’ and ‘Can I justify this change with subject knowledge or an accepted rule?’

Human oversight is essential because answers in Humanities can depend on nuance, scope and defensible interpretations. A teacher may refine or reject an AI suggestion. Students benefit from seeing that careful judgement rather than unquestioning acceptance is the goal.

Markly Does Not Replace Subject Knowledge

A response filled with confident sentences cannot compensate for an incorrect understanding of a river process, historical chronology or governance concept. Humanities learning still requires reading, note organisation, retrieval of accurate information and use of concepts in appropriate contexts.

Before revising an essay, ask whether the student knows the material well enough to improve the explanation. If not, review a reliable lesson or textbook section first. A writing assistant cannot supply valid evidence of comprehension when the student never had the underlying knowledge.

Practice should therefore alternate between learning the content and applying it. Reading, retrieval and discussion build the raw material; written responses test whether that material can be organised into an argument or explanation.

An Eight-Week Teaching Cycle for a Punggol Learner

Weeks 1–2: Collect a small sample of independent responses. Categorise recurring problems: inaccurate content, missing explanation, weak evidence, poor structure or language mechanics. Choose the most important one.

Weeks 3–4: Model the repair on a short example. Have the student explain what makes the revision more accurate. Use two different tasks that demand the same underlying reasoning skill.

Weeks 5–6: Reduce prompts. Ask the learner to identify their own weak step before reading comments. Give limited feedback and check whether the self-diagnosis agrees with the teacher’s judgement.

Weeks 7–8: Use a fresh question under ordinary time constraints. Compare it with the starting sample and note which skills have transferred. Adjust the next focus according to the evidence.

This is an illustrative learning routine, not an announced Markly rollout timetable or an official MOE eight-week programme.

Why the Punggol Context Matters

A learner at Edgefield, Greendale, Punggol or Yusof Ishak Secondary may be studying similar national subject frameworks but encountering different class tasks and school schedules. A parent cannot establish Markly participation from the school’s location in the same town.

What can be shared across households is the method of learning from feedback. Families can ask whether students understand comments, can distinguish content errors from expression errors and are able to retry without copying. These questions remain useful with handwritten comments, teacher discussions or a teacher-supervised digital workflow.

Punggol itself provides examples for thinking: urban runoff, shared transport, ageing neighbourhood infrastructure, migration narratives and community choices. Use such examples to practise reasoning, but follow the syllabus for formal examination content and do not assume a local project is an assessed topic.

Responsible Data Practices for Written Assignments

A school writing sample may mention identifiable peers, family events or sensitive experiences. Do not upload a student’s work to an unapproved public AI site merely because a story says MOE is experimenting with marking tools.

Use the school-approved assignment channel and follow the teacher’s instructions about what is permitted. When asking an outside person for guidance, remove names, class numbers, contact information and other identifying details from samples.

Discuss academic integrity as well. The child may be allowed to use a school tool for formative feedback while being expected to write the actual assessed response independently. Distinguish authorised revision assistance from submitting material that another system composed.

What Markly Does Not Prove About National Examinations

The September report concerned classroom formative feedback and pilots. It did not announce Markly as the marker of SEC examination scripts, a replacement for teachers or a change to national assessment requirements.

Punggol candidates preparing for the 2027 SEC should follow the applicable SEAB syllabus and their school’s examination guidance. A teacher may use new tools to provide practice feedback, but students still need subject knowledge and independent response skills.

When reading an article saying ‘AI is marking essays in Singapore schools’, ask which task is being discussed. Draft feedback in a lesson, a school-based assessment and a national examination are not interchangeable processes.

What We Still Do Not Know

The public September report did not publish a complete list of schools in the English, History and Social Studies pilots. It also did not supply a binding nationwide launch date for those subjects. An individual case study at Canberra Secondary does not establish deployment at a Punggol school.

We should also avoid inventing accuracy guarantees, time savings for every teacher or proof that students’ grades will rise. A classroom trial may suggest promising practices, but general adoption depends on evaluation and decisions that need their own evidence.

If MOE later publishes an updated programme status, a responsible guide should use the new notice and explain how the policy has changed rather than silently treating earlier projections as fact.

How to Ask a School About Its Plans

Ask which formative feedback tools teachers currently use for the child’s subject, whether any piloted tool is school-approved, and how teacher review works before comments reach students. Ask whether the child can discuss a confusing or questionable suggestion.

Ask how the student should revise. Is there an opportunity for a short independent retry? Does feedback target reasoning, content and evidence as well as grammar? Is the workload proportionate to the learning goal?

Do not demand that a school adopt Markly because a news article described a successful example elsewhere. Equally, do not dismiss a thoughtful teacher-led pilot purely because AI is involved. Evaluate the resulting quality of learning and teacher oversight.

A Parent’s Decision: What Matters Most?

There are three reasonable priorities. First, the child must understand the syllabus concepts and the task they are answering. Second, they must receive feedback that identifies real, manageable learning needs. Third, they must practise the repair independently until it becomes reliable.

If those conditions are met using ordinary teacher comments, that is good teaching. If a supervised tool makes the cycle faster or clearer, it can be useful. If it creates excessive revisions without understanding, the additional activity may not help.

Technology should be judged by educational results, not by how futuristic the interface looks. The presence of a new system at a school is neither proof of quality nor proof of harm.

Common Misconceptions

‘Markly is now required in every Punggol Secondary school.’ No public source reviewed for this guide establishes that. Geography availability and pilots in other subjects must be distinguished.

‘Markly is the same tool as AIMS.’ They are separately named systems with different publicly described uses.

‘The computer’s marks are final.’ MOE says teachers remain responsible for feedback. The reported classroom example also showed a teacher deliberately withholding machine grades during drafting.

‘Six drafts are the required number.’ The six-draft sequence was a case study, not a national rule.

‘An essay is correct if every sentence is grammatical.’ Humanities answers need valid evidence, concepts and explanations; English writing needs purpose and audience as well.

‘A new tool means a different SEC examination.’ The formative-feedback report did not change national examination requirements.

Frequently Asked Questions

What is Markly? It is an MOE-developed AI-assisted formative feedback tool for essays and extended written responses, described in a September 2026 report.

Which subjects were named? Geography was available to Secondary and JC teachers. English, Social Studies and History were being piloted in some Secondary schools.

Is Markly definitely used in my child’s Punggol school? That cannot be inferred from the national news. Ask the school directly.

Does Markly replace the teacher? No. MOE said teachers review, edit or reject suggestions and remain responsible for feedback given.

Can Markly give a grade? The reported example says the tool can provide grades, but one teacher chose not to use that feature during drafting. Do not assume every school’s practice is identical.

Does it work only with English essays? No. Geography and the other named Humanities subjects were central to the publicly described programme. Read the actual subject scope carefully.

Will History and Social Studies be available to all schools in 2027? The report described pilots, not a confirmed universal 2027 rollout.

Will it mark SEC papers? No role in national examination script marking was established by the reported formative-feedback programme.

What should a student do after getting comments? Identify the first important misunderstanding, revise it independently and try a new task that uses the same reasoning skill.

Can a student question an AI-assisted comment? Yes, respectfully, with evidence and teacher guidance.

Are more revisions always better? No. Revisions should improve understanding and independence rather than create repetitive polishing.

How can a parent help without knowing the subject deeply? Ask the child to explain what they changed, why it matters and whether they can apply the principle to a new question. Seek teacher clarification for technical disputes.

Will using AI automatically improve grades? No. Results depend on subject knowledge, quality of instruction, engagement and transfer to independent work.

Do families need to buy access? The reported programme concerns school teaching tools. Do not purchase an unofficial ‘Markly account’ on the assumption that it is required.

Sources and Reading Pathways

For the dated account of programme status and the described classroom example, read The Straits Times, 5 September 2026. For MOE’s broader AI literacy direction, consult the 2026 Committee of Supply announcements. Official syllabuses and examination instructions remain with SEAB and MOE.

Within this series, AIMS for Secondary Chinese explains the different September 2026 language marking initiative. The Punggol Education and Town-Life Index collects other school-stage guides. For a subject-independent example of diagnosis, repair and independent retry, see the Clementi Secondary 1 Mathematics learning guide; it is not a description of Markly.

The Future Is Feedback Students Can Use

Markly makes visible an important opportunity. If technology can assist teachers with part of the annotation workload, educators may have more room to discuss misunderstandings and guide repeated attempts. That opportunity becomes real only when students use the resulting comments to build knowledge and judgement.

A Punggol learner should be able to say more than ‘My essay was corrected.’ They should be able to say, ‘I misunderstood the relationship; now I can explain it,’ or ‘My evidence did not justify my inference; this new example does.’ When that happens on a fresh task without a machine or teacher supplying the answer, the purpose of feedback has been achieved.

Fact-checked for the status known on 23 October 2026. Geography use and selected subject pilots are distinguished from any unannounced universal rollout. Teaching scenarios in this guide are illustrative.