AI is used selectively
Premium does not mean “AI everywhere”. The site prefers authored questions and deterministic marking where a reliable rule or mark scheme can do the job. AI is used where explanation, judgement or personalised synthesis adds genuine learning value.
What may be sent to the AI service?
| Feature | Typical information sent |
|---|---|
| AI-assisted marking | The chemistry question, marking guidance/model material and your submitted answer. |
| Full Topic Review feedback | Your completed review evidence, marks/responses and relevant mastery or misconception information. |
| Exam feedback | Paper/question context, submitted answers, scores, assessment-objective/topic evidence and worked-answer context. |
| Dr Clay Tutor | Your current message, recent tutor conversation and relevant learning profile such as mastery, self-ratings, misconceptions and current learning actions. |
| Personalised practice | Relevant topic/skill evidence needed to choose or generate an appropriate learning task. |
What gets stored by this site?
The site may save your learning answer, tutor conversation, generated feedback or resulting learning evidence so that the feature works across visits and you do not need to regenerate the same feedback repeatedly. Separately, the AI usage meter records operational information such as feature, model, token counts, estimated cost and time. That usage log is not a second copy of the full prompt.
OpenAI processing
The current AI integration uses the OpenAI API. OpenAI states that API inputs and outputs are not used to train its models by default unless the API customer explicitly opts in to sharing. OpenAI's standard API data-retention rules may keep certain API content for a limited period for abuse monitoring, subject to the API service/configuration and legal requirements. We do not claim a special zero-retention configuration unless it has actually been enabled for this deployment.
Optional trusted web search
Dr Clay Tutor includes an optional “Use trusted web sources when helpful” switch. When you turn it on, the AI request may use web-search tooling restricted by the application to selected chemistry/science sources. Do not use that switch to submit personal information.
AI can be wrong
AI-generated feedback is advisory. It can misunderstand chemistry, miss a marking point or produce an inaccurate explanation. The site therefore keeps deterministic marking and authored worked answers as the source of truth where they are available, and AI feedback does not silently change a recorded mark.
No high-stakes automated decisions
The AI may prioritise revision or suggest what to practise next, but it does not award an official A-Level grade, decide university admission, make disciplinary decisions or make other comparable high-stakes decisions about a student.
Report a problem
If an AI response is wrong, inappropriate or unhelpful, use the in-site feedback tools. This helps improve prompts, question quality and safeguards without requiring direct private messaging with Dr Clay.