Sources and evidence

Every claim about efficacy on this site carries one of six tiers, and the tier is the only place efficacy is expressed. The order below is fixed and is the order things are presented in.

Established evidence

Supported by systematic review, meta-analysis or a regulatory or clinical position, from more than one independent original source.

Emerging research

Real human studies exist and point somewhere, but the body of evidence is small, mixed, short-term or heavily industry-funded.

Plausible mechanism, limited human evidence

There is a credible biological or chemical reason it might be true, and that is different from evidence that it is.

Traditional use

Long-standing use in a named culture or trade. Recorded as a practice, which is not a statement about whether it works.

Cultural belief

Widely held within a community and recorded as a belief. ChocolateHQ neither endorses nor corrects it.

Commonly claimed, but not supported

Frequently repeated and not supported by the evidence consulted. Recorded so that the correction is findable.

Reaching established evidence requires three things at once: at least one source of a qualifying type — a systematic review, meta-analysis, regulatory position, clinical guidance or food standard; at least two independent originals, counted by collapsing every republication and aggregator onto whatever it is downstream of; and every supporting source verified at the level of the passage being relied on.

The independence rule matters more here than it might elsewhere. “Dark chocolate is good for your heart” appears on tens of thousands of pages, almost all of which trace back to a small number of trials, many industry-funded. A catalogue that counts URLs counts marketing reach; one that collapses to distinct originals counts evidence.

The register holds 26 sources, of which 25 are external documents and 0 are verified at the passage level.

Every external source in the register is currently recorded as citation identity confirmed. That means the document is known to exist and its title, publisher, year and identifier are recorded correctly. It does not mean anyone has re-read the specific passage being relied on.

The consequence is deliberate and visible: because the top tier requires passage verification, no health topic on this site currently sits at established evidence. Two of them arguably belong there on the merits — the cocoa flavanol work and the theobromine toxicity in dogs. They are held one rung down until a human has done the retrieval.

That is the gate working rather than failing. A foundation that shipped with a full top tier would have proved only that the tier was cheap. Clearing it is the first item of outstanding work.

Separately from the tier ladder, individual facts carry their own confidence marker. Several of these are presentation states that exist so a page can say something true rather than quietly omitting a field:

• Recorded — established and attributed.
• Inferred — a reasonable reading of the available material rather than a direct statement in it. Most origin flavour claims sit here.
• Sources disagree — genuinely contested. Peru carries no country-level flavour claim for this reason.
• Not recorded by the sources consulted — the question was asked and the material did not answer it. This is why the Criollo production-share figure that circulates widely does not appear here.
• Not applicable — a structural fact about the product, such as white chocolate having no cocoa-solids percentage.

There is no numeric confidence score anywhere in this site and none will be added. A single number between zero and one invites false precision and conceals exactly the uncertainty the model exists to surface.

This site holds citation metadata and opaque locators — a section number, an annex reference. It does not store the text of other people’s documents, and its schema rejects any record that tries to.

It also does not print figures it cannot attribute. Where a number circulates widely without a traceable original — the share of world production that is Criollo, the exact cadmium limits in a regulation nobody here has re-read — the catalogue records that the figure is unavailable rather than repeating one.

Where a source’s funding is material it is recorded. In this domain that is not a formality: a substantial proportion of cocoa flavanol trials have been funded by confectionery manufacturers. That does not make them wrong. It is something a reader is entitled to know without having to look it up.

Editorial policy — what this site will and will not publish. Health and evidence — the topics this page governs.