Skip to content
QuranBench
Editorial

About QuranBench

You should not have to take anyone’s word for what the Quran says — including ours. QuranBench exists so you don’t have to. This page is for the sceptic: what this is, what it refuses to do, who is behind it, and how it stays honest.

What this is

QuranBench is an open research workbench for the Quran. Every Arabic word is a permanent, addressable object with its own page: its forms, its position, its root and morphology, every other place the same form occurs, and how translators rendered the verse around it. Search runs over the text by its own words — form, root, lemma, proximity — and every result shows the query and the corpus version that produced it, so you can reproduce it.

The promise is simple: do not accept our interpretation. Open the evidence, reproduce the search, and investigate the Quran yourself.

Why it exists

Most Quran sites ask you to trust a rendering. The interesting questions — where a root recurs, when different Arabic words are translated the same way, how editions disagree — are exactly the ones a page of prose hides. QuranBench treats those as computations you can run and check, not conclusions you must accept. It is built to be cited: stable URLs, versioned data, and a machine-readable API alongside every human page.

What it claims — and does not claim

It claims that every computed figure — a count, a frequency, a co-occurrence, a similarity score — is derived deterministically from a named corpus version and reproduces exactly. Where a fact comes from an outside source, it is labelled with that source.

It does not claim to settle meaning. The morphology is a third-party scholarly reconstruction (Leeds QAC) with its own editorial choices; the translations are human editions we attribute and never blend; our own investigations are labelled editorial and are the only layer that is our opinion. Disagreeing with any of these is legitimate, and every layer is tagged so you always know which one you are reading.

Above all, the Quranic text is never modified. Normalised, tashkeel-stripped and segmented forms are stored as separate, labelled fields; the source text is immutable and always attributed to Tanzil. Nothing we create is allowed to visually resemble scripture. See how this is built for the exact provenance of every field.

Who runs it

The maintainer has not yet decided whether to publish a legal name, so the identity here is deliberately left as a clearly-marked placeholder: NOT YET DISCLOSED. We would rather show a visible placeholder than invent a name or quietly omit the question — a project that asks for your scepticism should be scrutable about its own authorship. This line will be filled in, in one place, when that decision is made.

Regardless of the name attached, accountability does not rest on trust: contributions are governed by published contributor terms, moderation actions are logged, and anything you believe is wrong has a place to be reported.

How it is funded

QuranBench takes no money to show you anything. Nothing — no corpus, no search, no evidence, no download — is ever behind a login or a payment. There is no advertising and no tracking of what you read.

The funding model is still being settled: the maintainer has not yet finalised whether to operate under a charity or community-interest registration before accepting donations, a decision that affects tax treatment and must be made before any payments are wired. Until then the project is run at its own small cost. When donations open they will fund hosting and data work only, and a login will still gate nothing but a user’s own private work.

Editorial policy

Six provenance layers exist, and every rendered element carries exactly one: Quranic text (immutable, Tanzil), computed (derived from the corpus, reproducible), morphology (external annotation, Leeds QAC, GPL), translation (attributed human editions), editorial (our own investigations), and community (contributed responses). The tag is enforced in the component layer, not left to whoever wrote the page.

Investigations are structured, not free-form opinion: each is a falsifiable claim with a runnable query, pinned evidence, and its own required counter-evidence. Responses must declare a type and cite evidence. Corrections are reviewed by a person and, where they touch the data, ship as a new immutable corpus version rather than a silent edit — so every citation made against an older version still resolves.