What the AI is for
QuillSquare uses AI to analyse submitted English writing and produce structured educational feedback. The report organises assessed strengths, areas to develop, examples and suggested next steps. It is designed to support a writing conversation and future practice, not replace a parent, tutor or educator.
The first launch is primarily designed and marketed for England-oriented 10+, 11+ and 13+ entrance-writing preparation, with an upcoming-September 11+ default when no different supported context is supplied. These assessments do not follow one prescribed English course. The same core analysis can be used for general English-writing improvement.
Where AI is used
Providers, models and report structure can change. QuillSquare records technical version information needed to trace an evaluation without exposing the protected prompt or raw learner work in ordinary logs.
| Stage | What happens | Current provider role |
|---|---|---|
| Safety and input checks | Automated checks decide whether user-controlled text or files may proceed. Rejected content is stopped before evaluation. | Anthropic commercial API. |
| Document and handwriting reading | Images or PDFs are converted and text is extracted so the user can correct OCR errors. | Anthropic commercial API. |
| Writing evaluation | The supplied writing, task and permitted learner context are used to generate the structured report and next-step suggestions. | Anthropic commercial API. |
| Progress summaries | QuillSquare software compares completed evaluation data inside the authorised account relationship. | QuillSquare processing of generated reports; this is not general-model training. |
A responsible adult must review the feedback
AI output can be inaccurate, incomplete, inconsistent or biased. A responsible parent, tutor or educator should review the report before presenting it to a learner or relying on it.
- Check quotations and examples against the submitted writing.
- Use professional or parental judgement about tone, difficulty and next steps.
- Do not treat a score or band as an official exam result or guarantee of future performance.
- Do not use QuillSquare as the sole basis for a school, disciplinary, admissions, health, SEND or other significant decision.
- Do not assume a report proves authorship, cheating, plagiarism or whether AI was used by the writer.
SEND-aware evaluation
When a responsible adult selects a non-None SEND profile, the condition is sent as evaluation context so the AI can change the strictness, feedback wording and suggested next steps. This is not merely a display preference; it affects the report. The selection is optional and must not be used to diagnose a learner.
A parent or guardian must make the choice with explicit information, or a tutor must confirm that the parent or guardian explicitly agreed. The adult should choose only a listed condition genuinely relevant to the requested adaptation and should not add a medical history or diagnostic document.
Changing the profile later does not rewrite an earlier report. Each completed report is a point-in-time output based on the context supplied then. A valid full withdrawal or erasure request may require deleting the whole affected report rather than removing the SEND marker and leaving a misleading evaluation.
Automated checks are limited
QuillSquare’s moderation gate can reject content that appears unsafe, inappropriate, highly identifying or unsuitable for processing. It may reject safe work by mistake or fail to recognise concerning material. A rejection is a technical content decision, not a safeguarding assessment or finding about a real person.
QuillSquare does not routinely have a person read rejected or accepted writing. The gate does not alert the company founder, a parent, tutor, school, police, Childline or another service. A user can ask for product help with a suspected false rejection, but should not copy the raw writing into an email.
Training and product improvement
QuillSquare does not use learner writing to train its own AI model. It does not intentionally send learner material through consumer chat products or voluntary provider feedback/training tools. Providers process the minimum material needed for their role under their commercial service terms, including any limited security, abuse-prevention or legal retention those terms permit.
QuillSquare may use the generated evaluation history to provide account-linked progression, charts, comparisons and reports. If an adult reports a specific recognition or feedback fault, the company may inspect only the necessary protected source and report to investigate that complaint. Generic defect learning and synthetic examples are preferred for any later prompt or product improvement.
Appropriate educational use
- Use feedback after a piece has been written, or for practice and revision allowed by the responsible adult or setting.
- Do not use QuillSquare to answer a live, supervised or controlled assessment where outside help is prohibited.
- Do not submit AI wording as the learner’s own work where the applicable rules do not allow it.
- Do not use a learner profile or report to impersonate, shame, rank publicly or manipulate a child.
- Do not submit unnecessary personal, medical, safeguarding or confidential information.
Questioning a report or reporting a fault
A difference in educational judgement, score, tone or preferred wording does not by itself mean the report is faulty. If a report is blank, corrupt, unreadable, unrelated, substantially incomplete or materially different from the described service, use the refunds route. For another product concern, email help@quillsquare.app with the safe reference shown by QuillSquare.
Do not email learner writing, a raw report, a password, internal identifier or private guest-result link. QuillSquare does not promise routine human marking or review of every challenge.