How Menteez Uses AI

We use AI in exactly one place: building the question bank, before anyone sees it.

Here is the whole process:

• We define the standard first. We collect the up-to-date syllabus for each board, class and subject.
• AI drafts more questions against that standard. We ask an AI model to write candidate questions for a given board, class, subject, chapter and difficulty level. The request contains no information about any child. It says, in effect, "write questions for CBSE Class 5 Science, chapter 3, medium difficulty". It cannot say anything else, because it is never given anything else.
• Automated checks reject most of them. Every candidate goes through checks for things like duplicate questions, implausible wrong answers, circular explanations, questions that are off the syllabus, and answers that do not match the question.
• Senior teachers and an internal panel review what survives. Paid senior teachers and an internal panel review samples from every batch against the quality standard.
• Only approved questions reach your child.

We want to be precise, because this matters.

What is true: every published question has passed the automated checks and comes from a batch that teachers and an internal panel reviewed and approved.

What we do not claim: that a teacher has personally read every single question. Teachers review samples from each batch and reject whole batches that fall short. That is a real quality control, and it is not the same as individual sign-off. We would rather tell you that than let you assume otherwise.

So: some questions will still be wrong. A typo, an ambiguous option, an explanation that does not quite explain. When you find one, please tell us.

We are as clear about the boundaries as about the uses.

• No AI chats with your child. There is no tutor, no assistant, no conversational feature.
• No AI marks your child in a way that changes what they are shown next. Scoring is arithmetic: the answer matches the key, or it does not.
• No AI builds a profile of your child. Worksheets are matched to board, class and subject, not to a model's opinion of your child's ability.
• No automated decision that significantly affects your child. Nothing about your child's account, access, or results is decided by a model.
• No AI-generated content about your child is shown to you or anyone else.
• No AI is used for advertising or marketing to your child. We do not advertise to children at all.

This is the commitment we are most careful about.

The requests we send to an AI model contain only a board, a class, a subject, a chapter, and a difficulty level. They contain no names, no dates of birth, no schools, no answers, no scores, and no identifiers of any kind. There is no technical route by which a child's information reaches a model, because the information is never put into the request.

Nothing you or your child gives us is used to train, fine-tune, or improve any AI model, ours or anyone else's.

A person is accountable for what Menteez publishes, not a model.

• Teachers and an internal panel set the quality standard and make the approval decision.
• Batches are rejected by people, not by a score.
• Every published question can be traced back to the batch it came from and the review it passed.
• When we get it wrong, that is our responsibility, not the model's.

Being straightforward about what this technology does not do:

• Practice questions are not a syllabus. They support your child's schoolwork; they do not replace their teacher or their textbook.
• Coverage is uneven. Some chapters have more good questions than others.
• Difficulty labels are a judgement, not a measurement. What we call "medium" is anchored to teacher examples. Your child may disagree, and they may be right.
• A score is a snapshot. It reflects one worksheet on one day. Please do not read more into it than that, and please do not let your child read more into it either.

If we introduce any new AI feature, particularly anything that interacts with your child or uses their data, we will update this page before it launches, and we will ask for your consent again where the law requires it. We will not add such a feature quietly.