The anti-answer guardrail
A dedicated detector inspects every tutor reply and blocks rewritten sentences, model paragraphs and copyable text before the student sees it. Academic integrity isn't a policy page here — it's enforced in code.
Years 3–10 · ACARA v9 · NESA · VCAA · QCAA · SACE · NAPLAN-style
The only AI writing tutor that refuses to write for your child.
Every other tool hands over a better paragraph. Your child pastes it, the mark goes up, the skill doesn't. OwnMark Write marks against the real Australian rubric, names the one habit holding them back, and makes them fix it themselves — through six locked steps, in their own words.
- Never rewrites a sentence — guardrail-enforced
- Marked against ACARA work samples, not vibes
- Same draft, same mark — every time
- Parent gap report: year level, in years
20-minute baseline writing task · no credit card · gives you their working year level and gap in years
Year 7 · Narrative · Draft v2 · live
One habit · Sentence variety
Blocked: “Write it for me”
The tutor will not produce text a student can paste. Ever.
Ideas
4 → 6
Structure
3 → 5
Vocabulary
4 → 5
Illustrative report from three weeks of guided cycles.
For the parent reading this at 9pm
You don't want a better essay. You want a child who can write one without you.
You've read the same paragraph three times and still can't say what's wrong with it. The teacher's comment said “needs more detail”. The tutor costs $80 an hour and you still don't know if it's working.
And underneath all of it is the quiet worry: that they'll get through school leaning on something that thinks for them.
This is the tool that refuses. It sits beside them, asks the question a good teacher would ask, and waits. Every sentence stays theirs — and every month you get a plain report of exactly which year level they're writing at, and what moved.
AI didn't make writing easier to teach. It made it easier to fake.
What generic AI does
- Hands over an improved paragraph to paste.
- Scores a finished essay, then stops.
- Gives a different mark for the same draft twice.
- Uses generic US rubrics with no year-level ceiling.
What OwnMark Write does
- Asks a question the student has to answer in their own words.
- Coaches the whole process — brainstorm to reflection.
- Deterministic marking: same draft, same mark, quoted evidence.
- Judged against real Year 3–10 Australian work samples.
Four things nobody else does together
Plenty of tools mark essays. Plenty give feedback. The combination below is the product.
Process-locked six-step cycle
Brainstorm → write → self-review → hints → rewrite → reflect. Steps unlock in order, so students can't skip straight to feedback. The evidence of learning is the process, not the final paragraph.
Deterministic, evidence-checked marking
Temperature 0, fixed seed, retrieved rubric, and a chain of evidence → rubric mapping → gap → question → score. Every quote is verified against the student's actual draft; unverifiable marks are re-run and flagged for human review.
Australian curriculum lock
ACARA v9 achievement standards and work sample portfolios, with NESA, VCAA, QCAA and SACE overlays plus NAPLAN-style criteria. Feedback is capped to what's expected of that year — not an adult standard.
Try it: type a draft, watch it refuse to write
Paste a few sentences of your child's writing below. You'll see exactly what a session feels like — what the tutor notices, the question it asks back, and the one line it sends your child away to fix. No account, nothing saved.
What the tutor sends back
Noticing 1 of 2
4 sentences in a row start with “i”. Your reader starts to hear a drumbeat instead of a story.
What was happening around your character at that moment? If one sentence started with the place or the sound instead of “i”, which one would you choose?
Your turn: Rewrite just that one sentence so it begins with something other than the person doing the action.
Noticing 2 of 2
You lean on “very” and “really” to make the feeling bigger. Those words tell the reader how to feel rather than letting them feel it.
If you deleted that word entirely, what would you have to show instead — a hand, a breath, a sound — so the reader still gets there?
Your turn: Cross the intensifier out and replace it with one concrete detail of your own.
Notice what's missing: no rewritten sentence, no model paragraph, nothing to copy. Every reply is screened and blocked if it hands over usable text. In the full workbook this same exchange is graded against Australian work samples for your child's year.
One page. Six steps. Repeat until it sticks.
01
Brainstorm
The tutor questions the idea before a word is drafted — audience, purpose, the shape of the piece.
02
Write
A plain ruled page. No bold, no italics, no toolbar. Line numbers and word count only.
03
Self-review
The student marks their own draft against the year-level checklist first. Judgement before feedback.
04
Hint ladder
Four rungs: nudge, locate, name the rule, show a parallel example — never their sentence rewritten.
05
Rewrite
Version 2 in their own words. The tutor re-marks and shows exactly which criterion moved.
06
Reflect
What changed and why. The habit is logged, tracked across pages, and turned into the next drill.
Why the refusal is the feature
The research is unusually blunt about this
38%
of feedback interventions made performance worse, not better — chiefly when feedback pointed at the person or supplied the answer instead of the process.
Kluger & DeNisi, Psychological Bulletin (1996), meta-analysis of 607 effect sizes
d = 0.82
effect size for teaching writing as an explicit, self-regulated strategy — planning, drafting, revising — the single strongest lever in the adolescent writing literature.
Graham & Perin, Writing Next / Journal of Educational Psychology (2007)
2σ
the gap between classroom instruction and one-to-one mastery tutoring with corrective feedback. OwnMark Write is built to imitate the tutor, not the answer key.
Bloom, Educational Researcher (1984), the 2 Sigma Problem
Generation
students remember what they produce far better than what they read. When the AI supplies the sentence, the generation and retrieval effects are switched off entirely.
Slamecka & Graf (1978); Roediger & Karpicke, Psychological Science (2006)
Put together: feedback works when it targets the task and forces the learner to close the gap themselves. That is the entire design brief for this tutor — and the reason it will not write your child's sentence, no matter how they ask.
Everything in the workbook
Baseline diagnostic
A timed writing task returns a working year level and a gap in years — the honest starting line.
Skill & vocabulary ladders
Year-wise rungs for sentence craft, structure, cohesion and word choice. Progress is visible rung by rung.
Issue pattern tracker
Recurring weaknesses are logged into 12 canonical skill categories and converted into adaptive micro-drills.
Side-by-side exemplars
Below / satisfactory / exceeding samples for the student's year, so 'better' is something they can see.
Socratic tutor chat
Topic-locked to writing and curriculum. Off-topic, unsafe and adult content is filtered at three layers.
A–E grades explained
Criterion marks map to a year-relative A–E scale, with the descriptor for every score point shown.
Parent gap reports
Where they started, where they are, what moved this month, and the realistic timeline to at-standard.
Family accounts
Parents create and manage child logins, set limits, and see every cycle across all children.
Human review queue
Any mark whose evidence fails verification is held for a human to sign off. Trust is auditable.
Compared with the tools you're probably already paying for
We used each category the way a Year 7 family would: same narrative draft, same question — “make this better”. The difference isn't the marking. It's who does the thinking.
Best for real skill gain
OwnMark Write
Refuses to produce copyable text, locks the six-step process, marks to ACARA + state rubrics with verified evidence.
Best for a fast score
Essay-marking graders
Upload, get a band and comments in seconds. Useful before an exam — but the student learns nothing about how to revise.
Best for drilling grammar
Worksheet platforms
Endless practice on spelling, punctuation and comprehension. Strong coverage, almost no extended-writing coaching.
| Capability | General AI chatbot | Essay graders | Worksheet platforms | OwnMark Write |
|---|---|---|---|---|
| Refuses to produce copyable text | No | Partly | N/A | Enforced in code |
| Coaches the whole drafting process | No | Marks final drafts | No | Six locked steps |
| Australian year-level rubrics | Generic | Some exams | Curriculum-tagged | ACARA + 4 state authorities |
| Same draft → same mark | No | Varies | N/A | Temp 0 + seed + verification |
| Evidence quoted from the draft | Sometimes | Sometimes | N/A | Verified or re-marked |
| Extended writing, not multiple choice | Free text | Yes | Rarely | Every task |
| Longitudinal habit tracking | No | Score history | Topic mastery | 12-category skill trends |
| Parent-facing gap report | No | Rare | Usage reports | Year level + timeline |
| Child-safe, topic-locked tutor | No | N/A | Yes | Three filter layers |
The formula behind the difference
Every step is chosen to force cognitive work, not remove it
Retrieve before help
Generation & testing effect
The student drafts and self-marks before any feedback appears, so the knowledge is produced, not read.
Slamecka & Graf, 1978; Roediger & Karpicke, 2006
One habit at a time
Cognitive load management
Feedback is capped to one or two weaknesses per cycle, keeping working memory free for the actual revision.
Sweller, Cognitive Load Theory, 1988
Question, never answer
Task-level feedback
Feedback aimed at the task and the self-regulation layer helps; feedback that supplies the answer often hurts.
Hattie & Timperley, 2007; Kluger & DeNisi, 1996
Fade the scaffold
Zone of proximal development
The four-rung hint ladder gives the least help that still unblocks, then withdraws it as the skill firms up.
Wood, Bruner & Ross, 1976; Vygotsky
Tools that hand over the better paragraph shortcut all four. The child gets a mark; the skill stays exactly where it was.
Parents
You'll finally know whether the tutoring is working. A working year level, a gap in years, and monthly movement per criterion.
Students
A quiet ruled page and a tutor that treats you like a writer, not a prompt. One habit at a time, at 11pm if you want.
Teachers & tutors
Curriculum-grounded marking between sessions, class-level weakness patterns, and marks you can override before release.
Straight answers
Will the AI write my child's essay?+
No. A guardrail inspects every reply and blocks rewritten sentences, model paragraphs and anything pasteable. If a student asks it to write, it asks them a question back.
Which curriculum is the marking based on?+
Australian Curriculum v9 achievement standards and ACARA work sample portfolios, with NESA, VCAA, QCAA and SACE overlays and NAPLAN-style criterion marking.
How can an AI mark be trusted?+
Marking runs at temperature 0 with a fixed seed, retrieves the year rubric and calibration anchors, and must quote evidence from the draft. Quotes are verified; failures are re-marked and flagged for human sign-off.
What ages is it for?+
Years 3 to 10, across narrative, persuasive and informative writing, with the tutor's tone and vocabulary ladder set per year. Years 11–12 are planned.
Is it safe for children?+
The tutor is topic-locked to writing and curriculum work, with adult-content, violence and off-topic filters at three layers, and parent-managed child accounts.
Find out what year level your child is actually writing at — in 20 minutes.
One timed task. A criterion-by-criterion breakdown against the Australian standard for their year, and the first habit to fix. No credit card, no essay written for them.