Everything this site is built on, in one register: every live feed, dataset, scripture, scholarly work and image. Each entry names what it feeds, how it is accessed, and what is wrong with it. A site that grades other people's evidence has to publish its own.
Queried on a schedule through edge functions and cached. The site reads the cache, never the upstream.
Periodic releases, ingested and versioned. Figures are reproduced as published, never smoothed.
Population and adherence counts. Census-era figures and historical reconstructions are tiered separately and never presented with equal authority.
Named academic work behind the historical readings. Where scholars disagree, the disagreement is shown rather than resolved.
Primary religious texts, cited by chapter, verse or collection. Quoted verbatim; never paraphrased into a claim.
Every portrait on this site is a halftone derivative of a Public Domain or Creative Commons image hosted on Wikimedia Commons. No press, wire service or agency photograph is used, in original or derived form. Where a licence requires attribution, it is given here and in the card the image appears in. Derivatives of share-alike originals carry the same licence as the original. Entries with no acceptably licensed image render a geometric mark instead.
Instruments this one is measured against, cited and credited, never reproduced.
How demographic figures are ranked where two credible bodies disagree. Where no figure exists, the cell says so.
Post-1900 counts and surveys: Pew, national censuses, US Religion Census / ARDA, MuslimsInBritain.org.
1 AD–1900 scholarly reconstruction: Johnson / World Christian Encyclopedia, Stark, Bulliet, Vryonis.
A reconstruction built on partial census returns. Carries reconstruction weight, not census weight.
Pre-1 AD. Region shapes and a labelled range only — no numeric adherent figure is defensible.
A published forward model, not an observation. Dated to the model, not to the year.
As stated on the congregation plate, for the bodies whose counts it renders.
Granular denominational self-report aggregated to a continuous global series from AD 33.
KNOWN CRITICISM — Counts adherents generously; its China estimates run far above Pew's.
Transparent country-by-country composition estimates with published methodology.
KNOWN CRITICISM — The 2020 China methodology change — now formal zongjiao self-identification — creates discontinuities with Pew's own earlier figures. Sikhs, Jains and Bahá'ís are bundled into 'other religions'.
An exponential model at roughly 40% growth per decade.
KNOWN CRITICISM — Criticised for assuming a constant rate and for leaning on the Mormon growth analogy.
Onomastic conversion curves built from biographical dictionaries.
KNOWN CRITICISM — Small samples; assumes the first non-Arabic name in a lineage marks the convert.
Foundational reconstruction of Byzantine decline and Islamization from Ottoman registers.
KNOWN CRITICISM — The decline paradigm is contested — Peacock, De Nicola & Yıldız (2015) argue gradual syncretic conversion.
The standard annual Jewish demographic series.
KNOWN CRITICISM — The conservative 'core' definition differs materially from the 'enlarged' definition used by the JPPI. Both are rendered here; neither is averaged.
Full-population self-report at a fixed date.
KNOWN CRITICISM — Many omit religion entirely — the US census never has. China undercounts. Nigeria's 1973 religion census was suppressed over falsification allegations.
The most extensive US count: 372 bodies reporting congregations and adherents.
KNOWN CRITICISM — Denominational self-report; COVID hampered the 2020 non-Christian counts, which are likely undercounts.
The de facto UK mosque registry, privately maintained and landmarked.
KNOWN CRITICISM — Broad definition of 'mosque' — prayer rooms and shared spaces included. No official status; there is no ONS count.
Expert-crowdsourced coding of historical polities.
KNOWN CRITICISM — Coding subjectivity and uneven coverage across regions and periods.