How we verify every question

The pipeline

Every question on EndlessMed goes through the same process before a student ever sees it:

1. Generation, grounded in current guidelines.

   Questions are written using current evidence-based sources (UpToDate, First Aid, major society guidelines) as the required reference standard, not open-ended free generation.

2. A structural accuracy check.

   Before anything is cached, each question is checked for internal consistency — the stated correct answer must actually match the explanation given for it, lab values must include normal reference ranges, and answer choices are checked for realistic, non-throwaway distractors.

3. A separate AI quality-review pass.

   A second, separate model reviews the question before it is served to a student, and a question is only shown once it passes. Questions involving several competing clinical findings that must be weighed against each other get extra scrutiny at this stage, since that's where mistakes are most likely. To be clear about what this is: it's an independent AI quality review working from current guideline knowledge — not a human checking each question line by line against source documents.

None of this makes the process infallible — it's still AI-generated content, and no verification pipeline catches everything. What it means is that nothing reaches you as a single, unchecked AI response.

Report a question, and it gets a real second look

Periodic manual audits

Beyond the automated pipeline, generated questions are periodically spot-checked by hand against current clinical literature and reasoning — the same way a question bank's own question-writers would review their work. This isn't a claim that every question has been personally reviewed before it's served; it's an ongoing practice alongside the automated verification above, not a substitute for it.