Skip to content

The instrument

What marks your essay

Not a chatbot with a prompt. A versioned marking engine that we change deliberately, test against a frozen benchmark, and publish the history of — so a score means the same thing this week as it did last week, and you can see for yourself when it does not.

Marking since 29 May 2026 · current release 1.2

Every essay, every plan

Two AI Judges read your essay independently and never see each other's mark. Their scores are reconciled into one calibrated result, and when they disagree widely a third reading is called in to settle the scores.

Judge 1Judge 2

The deep mark

The Supreme Judge is a stricter AI review, not a person: two senior AI Judges from two different providers mark separately, each reading through a different specialist lens, and the Supreme Judge reconciles them into one Verdict — with a paragraph-by-paragraph breakdown and a worked rewrite.

Senior Judge 1Senior Judge 2

Release history

Every release is dated and says what it changed for you. We only move the number when the marking itself moves — a version that ticks for its own sake is noise.

  1. Examiner Engine 1.2+0.1 from 1.1marking now2026-10-01
    • BothWhen both Judges finish, each now gives every sentence of your essay a verdict — works, could be sharper, or costs marks — with one blunt line each. Your report shows the most important of them. Harsher words never change the mark.
    • StandardYour mark now weighs ACER's two published criteria equally: the thinking in your essay, and how you organise and express it. Before, organisation and expression carried two-thirds of the mark.
    • StandardMarking now takes a little longer — your report arrives within minutes. You can leave the page while your Judges mark; your report will be waiting.
    • StandardThe Judges now place your essay by what it does, not by how common a mark is, and they have worked examples at the very top of the scale.
    • StandardYour mark is now worked out differently, so a mark from this version and one from an earlier version are not directly comparable — the same essay can score differently. Compare essays marked by the same version.
    • StandardThe Judges no longer mark against your overall GAMSAT target. They coach you toward the next band up.
    • StandardA prompt you paste now asks whether it is Task A or Task B, so a reflection is always read as a reflection.
    • StandardWhen you revise, each Judge checks your change against the exact request you chose, and a revision verdict appears only when both Judges agree.
    • UltraThe Supreme Judge now deliberates with two senior Judges and rules once both are in — usually within the hour if you keep the page open, or up to a day or two if you close it — and we tell you when your Verdict is ready.
  2. Examiner Engine 1.12026-09-11
    • StandardOne of the two Judges was upgraded to a newer generation. The rubric, the scale and how long a mark takes are unchanged.
    • UltraNo change to the Supreme Judge this release — the same three readings, seated the same way.
  3. Examiner Engine 1.02026-08-16
    • BothTwo Judges now mark every essay, on every plan — not just paid ones.
    • BothA mark that loses a Judge to an outage is re-marked automatically, instead of standing as a single opinion.
    • StandardSharper reading of Task B reflective writing, and stricter task-response checking.
    • UltraOne of the Supreme Judge's three readings was upgraded — closer reading of argument structure on the deep mark.
    • UltraA short-handed Supreme Judge review is now stated on the verdict itself rather than only in the score.

    The first NAMED release — not the first marking. Essays marked before this are not back-dated into a numbering that did not exist at the time; they simply say “Examiner Engine”, with no version. What came before is below.

The interview engine

A separate engine with its own release line, because reading an interview station is a different claim from marking an essay.

  1. Examiner Engine — Interview 1.0two readers2026-09-12
    • InterviewTwo readers now read every interview station against Aptavia's published instrument, with receipts quoted from the candidate's own words.

    Two readers read the same answer independently and never see each other's reading. Where they disagree by more than one anchor, no number is published for that station — their receipts are shown side by side instead. Agreement between the readers has not yet been measured.

Before 1.0

Two-judge marking has run since 29 May 2026. These months predate the version numbering and are deliberately left unnumbered — the dates and the changes are real, and inventing versions to fill the gap would not be.

  1. 2026-08-14Standard

    Both Judges upgraded again, at the highest reasoning effort their generation allows.

  2. 2026-08-02Ultra

    The Supreme Judge is recomposed around maximum-intelligence AI readers from three different providers.

  3. 2026-07-23Both

    The reasoning seats behind question review and figure checking upgraded with them.

  4. 2026-07-22Standard

    Both Judges upgraded to a faster generation — same marking, less waiting.

  5. 2026-05-29Both

    Two independent Judges start marking every Section II essay, and their scores are reconciled into one.

See it mark your own essay

Your first Section II essay, marked free by both Judges within minutes — a scaled score, the rubric breakdown, and line-by-line annotations. No card.

Mark my essay free

Scores are calibrated estimates from our own rubric, built from ACER's two published assessment criteria. ACER publishes no Section II marking rubric and does not endorse this product; only ACER issues an official mark.

Each Judge has its own lean: one tends to mark a little higher than the other. We do not correct either one. Combining the two marks largely cancels it out, and our scale was set on the combined marks as they are. If a mark rests on one Judge alone, it carries that Judge's lean, and the mark says it was not cross-checked.