Verified against Perplexity Pro · 2026-08-06
Check a health or treatment claim against clinical evidence, not anecdote or marketing
An evidence-tiering prompt — explicitly not medical advice — that requires a health or treatment claim to be checked against clinical-trial evidence specifically, with anecdotal and marketing-sourced claims flagged and ranked separately rather than blended into one confidence level.
The prompt
Ready to copy — highlighted parts are example details you can swap.
Check this health or treatment claim against the actual clinical evidence. This is a research aid to help me understand what the evidence says and ask my doctor better questions — it is not medical advice, and any decision about my own health goes through an actual clinician, not this answer.
CLAIM
Magnesium glycinate supplementation before bed significantly improves sleep quality in adults with mild insomnia.
WHERE I ENCOUNTERED IT
A wellness influencer's Instagram post promoting a specific supplement brand
RELEVANT CONTEXT
Generally healthy adult, no diagnosed sleep disorder beyond occasional trouble falling asleep
EVIDENCE TIERS — do not blend these together
1. Systematic reviews and meta-analyses on this specific claim, if any exist.
2. Individual randomized controlled trials, noting sample size and whether it was tested on a population comparable to the context I gave above.
3. Observational studies or smaller/preliminary trials — genuinely informative but weaker evidence, and should be labeled as such rather than presented with the same confidence as tier 1 or 2.
4. Anecdotal reports, forum discussion, or individual practitioner opinion — potentially worth mentioning as context on how the claim is perceived, but explicitly not evidence for whether the claim is true.
FOR EACH TIER FOUND
State what the evidence actually shows, including effect size where reported, not just direction ("helped" vs. "reduced symptom severity by roughly X in Y% of participants"). Note any significant limitation the evidence itself acknowledges — small sample, short follow-up period, funding source, population that may not generalize to my context.
WEIGHING IT ALL
State an overall sense of how strong the evidence is for this specific claim, driven mainly by the highest tier of evidence that actually exists — a single tier-4 anecdote and a well-powered tier-2 trial should never produce the same confidence level in the final answer, even if you mention both.
If credible clinical evidence directly contradicts the claim, or if there is currently no credible evidence either way, say that plainly rather than defaulting to a noncommittal "more research is needed" that avoids stating what's actually known so far.
Close with an explicit reminder that this is a summary of publicly available research, not a diagnosis or treatment recommendation, and that any decision belongs with a licensed clinician who knows my actual medical history.Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
Health claims circulating outside clinical literature routinely mix genuinely rigorous evidence with anecdote and marketing copy in a way that reads as uniformly confident regardless of which one it actually is — a supplement brand's page and a peer-reviewed meta-analysis can both state a benefit in the same declarative sentence structure, and a synthesis that doesn't explicitly separate evidence tiers will tend to inherit that flattening rather than correct for it. Requiring a distinct tier structure, and explicitly forbidding a single anecdote and a well-powered trial from producing the same confidence level, is the direct fix for that specific failure — it forces the model's stated confidence to actually track the strength of what it found rather than the volume or persuasiveness of how the claim is usually phrased online, since marketing content is by nature written to be maximally convincing regardless of the underlying evidence quality. Asking for effect size rather than just direction matters because "helped" or "improved" conveys almost no usable information on its own — a treatment that improves an outcome for 5% of a narrow subgroup and one that improves it for 60% of a general population are both technically "shown to help," and only the effect size actually distinguishes a meaningful clinical finding from a statistically detectable but practically negligible one. And the explicit not-medical-advice framing, restated at the close rather than only as a disclaimer at the top, matters because a well-organized, evidence-tiered answer can read as more authoritative than a typical AI response precisely because of how carefully it's structured — which makes the reminder that this supports rather than replaces a clinician's judgment more necessary here, not less, since the format's own credibility is part of what needs to be kept in check.
Verified against
Perplexity Pro Sonar Pro · 2026-08-06
Changelog
- 2026-08-06 — Initial publish, verified against Perplexity Pro Sonar Pro and Academic focus mode.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
