Safety & Responsible Reference Use

Appropriate Uses

• Medical literature search
• Guideline lookup
• Professional reference
• Medical education
• Research orientation
• Journal club preparation
• Evidence comparison
• Scientific background reading

Not Appropriate Uses

• Emergency advice
• Autonomous diagnosis
• Autonomous prescribing
• Patient-specific treatment decisions
• Replacing professional judgement
• Replacing peer review
• Replacing systematic review methodology

Human Oversight

Healthcare professionals remain fully responsible for all clinical decisions. MedCite outputs must be verified against primary sources, local clinical guidelines, and institutional protocols before application in any clinical context. The Service is designed to support, not replace, professional medical judgement.

Built-in Safety Features

• Patient-specific query detection — queries describing individual patients trigger evidence-only responses with explicit disclaimers
• Emergency detection — potential emergency queries are flagged with guidance to contact local emergency services
• Prescribing guardrails — dosing queries receive guideline-framed information only, never patient-specific instructions
• Citation discipline — every factual claim must be supported by a cited source
• Source verification — users can click through to original sources for independent verification

Regulatory Context

The EU AI Act (in force since 1 August 2024) classifies AI systems intended for medical purposes as high-risk when they fall under medical device legislation. MedCite is intentionally designed as an evidence retrieval and summarisation tool, not as a medical device or clinical decision support system, to avoid this classification. Our intended purpose, product language, and technical behaviour are all aligned with this positioning.

Clinical Evidence Planner

MedCite uses a Clinical Evidence Planner that transforms opaque AI search into an auditable decision pathway. Every query is classified into a clinical task (e.g., drug timing, treatment guideline, diagnosis), entities are extracted (drug, condition, procedure), and evidence retrieval is planned accordingly. The system exposes this reasoning so clinicians can inspect why sources were selected, validate the task interpretation, and understand the evidence basis for each answer.

Key Features

• Transparent task classification — see how MedCite interpreted your clinical question
• Entity extraction — drugs, conditions, and procedures are identified and normalised
• Task-driven retrieval — searches use answer-bearing concepts, not just topic keywords
• Answerability validation — evidence packs are checked for required clinical concepts before synthesis
• Deterministic fallback — known clinical patterns trigger structured query templates as rescue path
• Source-role prioritisation — guidelines and systematic reviews are boosted over case reports and imaging-only papers
• EU AI Act–aligned transparency — audit trail exports task, entities, sources, and validation results