How we cite
Research briefs written with AI assistance have a specific failure mode: real paper titles attached to fabricated identifiers, or a DOI that resolves to a completely different paper. We hit it ourselves in August 2026 and built this rule set in response.
The rules
- Every citation resolves. Each DOI is checked against the Crossref API and the returned title is compared to what we wrote, not just the HTTP status. Each PubMed ID is checked against NCBI E-utilities.
- Numbers come from abstracts. Sample sizes, effect sizes, p-values, and age ranges are copied from the fetched abstract, not recalled.
- Primary sources only for claims about companies. Competitor privacy practices are quoted from the competitor's own published policy, read on the date stated. Business-practice claims cite public complaint records or reporting, not our own summaries.
- "Could not verify" is a result. If a source cannot be reached or the title does not match, the claim is removed or marked, never left in place.
- Every brief says where the evidence stops. The section is mandatory. A brief without one is not published.
What "verified" does not mean
It does not mean we agree with a study's conclusions, that the study is high quality, or that its result transfers to our product. It means the citation is real and the numbers are the study's own. Quality and transfer are discussed in the brief text.
Reporting an error
If a citation on this site is wrong, email partnerships@minischolars.co with the page and the line. Corrections are logged in the changelog.