Skip to main content

AI transparency

AI change log

Which AI model we run on, what each AI feature sends, and a dated record of every change to AI features, safety and privacy controls. Material changes are also announced by email and on the Account page.

The model we use now

Model
google/gemini-2.5-flash
Provider
Google (Gemini), reached through the Lovable AI Gateway
Processed in
United States
Training on your data
None — nothing you submit is used to train any model.
As at
2 September 2026

A change of model, provider or processing country is logged below and announced with at least 30 days' notice, as promised in the Privacy Policy.

AI features available today

Every one of these runs only when you press its button, and every one can be turned off under Account → AI features.

FeatureWhat it doesWhat is sent to the model
Essay markingSuggests a mark and criterion feedback against the QCE English rubric.Your essay text.
Answer and flashcard checkingCompares your written answer with the model answer and suggests a mark. Off until you switch it on.Your written answer.
Brain-dump gap suggestionsCompares what you recalled with the knowledge point and lists what you missed.Your brain-dump text.
HintsOffers a graded hint for a question you're stuck on.The question and how far you've got.
Handwriting scanningTranscribes a photo of your written working so it can be marked.The photo, with camera metadata removed first.

Changes

  1. Safety

    This change log, inline accuracy reminders, metadata stripping

    Every AI response now carries a one-line reminder that AI can be wrong. Photos you upload for handwriting scanning have camera metadata (location, device, time) removed in your browser before upload. This page becomes the record of AI changes.

  2. Privacy

    Rate & report, input notices, an AI features switch

    Thumbs up/down and a Report link under every AI response; a privacy reminder beside every AI input; Account → AI features turns every AI entry point off; a How our AI works guide; a published sub-processor register naming the AI provider.

  3. Feature

    Brain-dump gap suggestions

    After a brain dump on a knowledge point, AI can compare your recall with the knowledge point and list what you missed. Opt-in per use; costs credits.

  4. Feature

    Flashcards: opt-in AI answer check in Write mode

    Write mode can send a non-exact answer to AI for a judgement. Off by default; a switch in the session turns it on; one credit per check.

  5. Feature

    QCE English essay marking

    AI-suggested marks and criterion feedback against the 45-mark external assessment rubric, shown beside your own self-marking so you can compare.

  6. Feature

    Hint ladder and criterion marking

    Graded hints for questions, and AI-suggested criterion marks for written answers.

  7. Model

    Monthly AI credit allowance; Gemini 2.5 Flash

    AI features moved to a monthly credit allowance per plan. All features run on google/gemini-2.5-flash through the Lovable AI Gateway.

How you'll hear about changes

  • This page is updated the day a change goes live.
  • The Account → AI features card shows the most recent change.
  • Changes that affect how your personal information is handled are emailed to account holders in advance, and you're asked to review and accept the updated policy before continuing.
  • Questions or concerns: use Report this response under any AI output, or read How our AI works.