LEELALUMethodology quality
How each position's evidence was gathered and weighed. A rigorous review beats a confident summary.
Evidence method
How a sentence earns its place.
The method behind every claim, in plain language. Before any sentence of guidance reaches a family, it passes the rules on this page. If it cannot, it does not ship.
01 · Sources
The evidence hierarchy.
Not all sources are equal, so we rank them. These are the five tiers we accept, in order of preference. The ranking tells us where to start reading; it never settles a disagreement by itself. How we handle disagreements is section 02.
-
Government health agencies
Bodies like the CDC, NICHD, and NHS. Public mandate, published methods, and a duty to correct their own record.
-
Professional medical associations
Organizations like the AAP and ACOG, whose guidance is written and revised by practicing specialists.
-
Systematic reviews and major peer-reviewed guidance
Work that weighs many studies together instead of leaning on a single result.
-
Reputable academic medical centers
Patient education from institutions that put their name and reputation behind every page.
-
Expert editorial synthesis with disclosed limitations
Careful summaries by named editorial teams, used only when they state plainly what they do not know.
Never random blogs, affiliate content, forums, or model memory. Those are not evidence here.
02 · Disagreements
When reviewed sources disagree.
Careful organizations sometimes reach different conclusions, or update the same conclusion at different speeds. When that happens, we adjudicate on seven factors, never by tier alone.
Recency
Which document reflects the current state of the evidence. A revision that supersedes older guidance carries that weight.
Directness of evidence
Whether the source answers the exact question, or only one near it.
Population match
Whether the studied group matches the families the claim serves: stage, age, and health context.
Jurisdiction
Whose guidance governs where our families live. At launch, the United States.
Consistency across bodies
Whether independent organizations converge on the same position, each from its own evidence review.
Clinical significance
Whether the difference would change what a careful parent actually does. Small wording gaps do not justify big editorial calls.
Every adjudication is written down. The claim record keeps the rationale: what disagreed, which factors decided it, and why the published wording won. A decision we cannot explain is a decision we do not ship.
Who decides today, honestly. Right now the accountable decision is made by the LEELALU editorial team. We have not yet contracted a clinical reviewer, so every adjudication is flagged for that reviewer to re-examine once one is on contract. We say this plainly instead of implying a clinician signed off.
High-risk conflicts are not published until resolved. If a disagreement touches safety and we cannot resolve it with confidence, the claim stays out of the library. Until it is resolved, the library simply has no answer there, and Lu says so.
03 · The claim record
What every claim record carries.
This is the registry standard we are building toward. There are 98 reviewed claims in the registry as of this build, each with a named source organization and an exact link; the remaining fields ship with the launch gates, and we say so instead of pretending the registry is finished.
| Field | What it means in practice |
|---|---|
| Claim ID | A stable identifier, so every sentence can be traced, audited, and corrected. |
| Exact approved wording | The sentence ships verbatim. No paraphrase or improvisation reaches a family. |
| Life stage and topic | The pregnancy weeks or baby months, and the subject, the claim applies to. |
| Source IDs | The specific documents behind the claim, linked exactly, never a homepage. |
| Evidence strength | An honest grade of how solid the underlying evidence is. |
| Reviewer | Who checked the claim against its source, on record. |
| Review date | When that check last happened. |
| Expiration date | A review-by date. Past it, the claim is re-checked or retired, not left to age. |
| Jurisdiction | The region whose guidance applies. At launch, the United States. |
| Allowed surfaces | Where the claim may appear, such as a Today card or an Ask Lu answer, and nowhere else. |
| Prohibited extrapolations | What the claim must never be stretched to mean, written down before anyone is tempted. |
04 · Refusals
What we will not do.
Some credibility shortcuts are common in this industry. We name them here so you can hold us to the refusal.
No vanity percentages
We do not publish an accuracy or satisfaction percentage without a published methodology behind it. A number without a method is decoration.
No schema shortcuts
We do not use medical schema markup as an authority shortcut for search engines. Our review status stays labeled as an organization, honestly.
No borrowed outcomes
We do not claim LEELALU improves outcomes without product-specific evidence. The product is a preview today, and we describe it as one.
05 · Corrections
When we get something wrong.
We fix it in the open: the correction is made, dated, and left visible rather than quietly edited away. Our correction process and editorial rules live at leelalu.com/editorial-standards.html.
Next step
This method is one part of a larger standard.
Evidence, AI limits, privacy, security, human review, and money, all stated as operational facts.