AI Claim Analyzer

No data is sent to Healthcare AI Learning — the analysis runs entirely in your own AI assistant. The result is an educational decision aid, not clinical, legal or regulatory advice, and not a procurement decision.

1. 1

   Get the Healthcare AI Learning Evaluation Guide

   A concise framework file you upload into your AI assistant alongside the prompt.

   Download Evaluation Guide (v0.4)
2. 2

   Copy the Quick Assessment prompt

   Both modes use the same Healthcare AI Learning framework. Quick Assessment compresses the output into a concise decision brief; Deep Evidence Audit returns the full structured audit.

   Quick AssessmentDeep Evidence Audit

   Copy Quick Assessment prompt

   Use the attached Healthcare AI Learning Evaluation Guide v0.4 as the analytical framework for this assessment.
   
   You are evaluating a healthcare AI claim, product, study, vendor proposition or pitch. Be critical, evidence-based and balanced. Do not assume the claim is either true or misleading. Determine exactly what the available information supports.
   
   IMPORTANT: Perform the full Evaluation Guide v0.4 analysis internally, but only expose the highest-signal, decision-changing findings in your output. This output contract overrides the guide's reporting/presentation structure, NOT its analytical framework or rigor.
   
   STRICT OUTPUT RULES FOR QUICK MODE:
   - Hard maximum: 400 words excluding the Sources section. If your draft exceeds 400 words (excluding sources), compress it before answering.
   - Do NOT reproduce the Evaluation Guide structure.
   - Do NOT provide a claim-by-claim audit.
   - Do NOT add extra sections.
   - Do NOT repeat findings.
   - If a detail does not change the decision, omit it.
   - Remove detailed metric/regulatory/safety/economic exposition unless it directly changes the recommendation.
   - Preserve the distinction between "not demonstrated" and "demonstrated not to work".
   - Distinguish regulatory status from evidence of effectiveness.
   - Do not infer evidence that is not provided. Label missing information as unknown.
   
   OUTPUT STRUCTURE — use exactly these sections and no others:
   
   1. **Verdict** — 2–3 sentences max.
   2. **Evidence that matters** — max 3 bullets total.
   3. **What is still unproven / concerning** — max 3 bullets total.
   4. **Three questions to ask** — exactly 3 numbered questions.
   5. **Recommendation** — choose ONE: WATCH / INVESTIGATE FURTHER / PILOT / CONSIDER ADOPTION / INSUFFICIENT EVIDENCE, then max 2 sentences of rationale.
   6. **Sources** — max 3–5 sources, prioritizing regulators, peer-reviewed studies, trial registries, and genuinely independent evaluations.
   
   OUTPUT FOOTER — end your response with exactly this line:
   "This assessment is an educational decision aid produced by an AI assistant. It is not clinical, legal or regulatory advice, and it is not a procurement decision."
   
   CITATION RULE:
   - Cite only sources you can name precisely and verify. Give the title, publisher/journal and year.
   - If you cannot verify a source, say so explicitly rather than inventing, guessing or approximating a citation.
   - Never present an unverifiable reference as if it were checked.
   
   IF THE GUIDE IS MISSING:
   - If the Healthcare AI Learning Evaluation Guide v0.4 has not been provided to you, stop and say you cannot complete this assessment without it. Do not substitute your own framework.
   
   Material to assess:
   [PASTE CLAIM / STUDY / VENDOR MATERIAL HERE]

3. 3

   Add the material you want to assess and run the analysis

   Paste the claim, study text, product page or pitch into the placeholder at the end of the prompt, attach the guide, and send.

The AI Claim Analyzer is a structured decision aid for education. It is not clinical, legal, regulatory or statistical advice, and it does not replace professional review.