Welcome to the Atlas.
One written text; a thousand years of careful notes on exactly what varies in it and what does not. This atlas turns those notes into something you can open and see — every claim cited to its source. Here is everything inside, and where to start.
The sections
How the verses are counted.
The Quran's text is one. Its verse numbering is not: the classical scholars of six cities counted the āyāt in slightly different ways, all preserved, all documented. Here they are, side by side.
The counting schools
Seven reckonings recorded in the classical ʿadad literature. The totals differ only in where a verse boundary falls — never in the words. The Kūfan reckoning is the one printed in most muṣḥafs today, and is the per-surah baseline used below.
Counted straight from the six riwāyāt computed · reproducible
The totals above are the classical figures. We also carry the full verse-by-verse text of six riwāyāt — and each one numbers its own āyāt. So the counts here are computed directly from that text, with no transcription in between. Same-city readings agree — Warsh and Qālūn (Madīnah) both total 6214; Qunbul and al-Bazzī (Mecca) both total 6220 — the regional ʿadad reckoning showing through the raw data. Pick a sūra to see each reading's āyah count; any difference from Ḥafṣ is highlighted.
Structural divisions documented
Verse-numbering is one way to cut the text. The tradition carries several others — some fixed and universal, some edition-specific. A few (rukūʿ, pages, the number of sajdas) genuinely vary by convention, and are marked as such.
Browse every sūra canonical counts
Look up any sūra to see how many āyāt each reading gives it and whether the counting schools are documented to diverge there — then bookmark it. The number on each card is the Kūfan (Ḥafṣ) baseline.
The Counting Lab.
A count is never just a fact about the text — it is a fact about the text plus a convention. This instrument makes the convention visible, so any numerical claim can be reproduced and stress-tested in the open.
1 · Verse totals, honestly
The single most-cited number about the Quran — how many verses it has — is not one number. Across the documented schools it spans a range.
2 · Adjust the variables — the full catalogue
Every documented choice that can change a count, gathered in one place. Pick a base reckoning, then toggle any convention on or off and watch the whole-Qurʾān verse total move. Each one is a real, catalogued variation — never an edit to the wording. This is the instrument the 74:30 work was missing: see exactly which conventions a number depends on.
Verse totals are the commonly-cited classical figures (al-Dānī and the regional traditions; sources vary by ±1–2). The word- and letter-level conventions change those counts rather than the verse total — see them live in the letter & word counter below. documented
3 · The freedom, measured
Enter a divisor and watch how many of the seven documented verse totals it divides evenly. This is the exact freedom the 74:30 audit quantified — a pattern can be genuinely present and still hang entirely on which convention you pick.
4 · Letters, and the freedom in a count
Count letters, words and verses over any single verse, a whole sūra, or the whole Qurʾān — in any orthographic edition — and watch every number move as you change what counts and how. This is how a letter- or word-based claim is built, and exactly how sensitive it is to convention.
5 · Reproduce & stress-test a claim
Numerical claims are documented here as claims — reproduced faithfully, then tested against the freedom the documented variables allow. A claim is never allowed to edit the text.
Rashad Khalifa · remove Sūrat al-Tawbah 9:128–129
Claim (1970s–80s): the last two verses of Sūrah 9 are not part of the Quran and should be removed, because they break a proposed 19-based count. Rejected by mainstream Sunnī and Shīʿī scholarship; unsupported by any manuscript, qirāʾa, or ʿadad school — every documented textual tradition includes them.
The documented record points the other way: Ibn Kathīr, citing the Ṣaḥīḥ, notes these closing verses were the written attestation Zayd b. Thābit located with Khuzaymah b. Thābit during the compilation, and reports them among the last revealed — a note about collection, not doubt. documented
The Basmalah has 19 letters
The starting point of the whole "Code 19" literature. Reproducible from the grounded text — and, as the letter counter above shows, true under one specific counting convention and not others. That fragility is the point: use the counter to see which convention yields 19.
Word count = muqattaʿāt-letter count = 39,349
The 74:30 Project's crown finding — and the one that survived the adversarial audit the project deliberately ran on itself. No outside critic forced it; the researcher stress-tested the claims as harshly as any skeptic would, out of a commitment to be fully objective and transparent. Across the twenty-nine disconnected-letter chapters, one number — 39,349 — is reached two independent ways: it is their total word count, and it is the tally of the muqattaʿāt letters within them — each chapter's own opening letters, counted in its own text and summed across the thirteen letter-groups. It factors as 19² × 109 (nineteen squared times the 29th prime). The word count doesn't depend on how you count the alif, but the letter tally does — so requiring the two to agree is a far stricter test than counting letters alone.
Held to its own rules, the project measured what survives once the fitting freedom is accounted for: a residue of genuine surprise on the order of 1-in-80 to 1-in-400 — suggestive, unexplained, and honestly short of miracle-grade. It rides on specific edition choices, which the atlas will let you vary. Read the story → documented finding
6 · Pattern search — brute-force the numbers
Sweep the documented numbers of the Qurʾān — every reading's verse, word and letter totals, per chapter and in the famous subsets — and see which divide evenly by a number you pick. The honest twist is built in: finding patterns is trivial. About 1 in 19 numbers is divisible by 19 by chance, so with enough of them you always find some. This shows the hits and how many chance predicts, together.
Each bar is the share of the same numbers divisible by that divisor. If 19 is no taller than its neighbours, it is doing exactly what chance does.
7 · Test a pattern — the adversarial null
Have a pattern you think is real? Reproduce it here, then let the tool run the audit the 74:30 Project ran on itself: it recomputes your quantity across every documented reading and reports how many also hit. Hold under a handful of settings and it is notable; hold under most of them and it was fitted. That fraction is the whole difference between a signal and a coincidence.
8 · The public patterns log
Document a pattern for the record. Submissions are held for review, then published here with their adversarial result — fitted or surviving — so the whole thing stays a shared, falsifiable record rather than a private claim. Run a test above, then add it below.
9 · The muqattaʿāt-letter test — reproduce 39,349
The actual claim, exactly. Each of the 29 marked chapters opens with disconnected letters; you count each chapter's own opening letters within it. The project's documented scheme gives 13 group totals, every one divisible by 19, summing to 39,349 = 19² × 109 — the same number as the (edition-robust) word count of those chapters. Below, the documented tally sits beside a plain Ḥafṣ count, so you can see exactly where the number is solid and where it leans on the counting convention.
The word count (39,349) reproduces in any edition. The letter tally reaches 39,349 with all 13 groups divisible by 19 only under the project's specific per-sūra edition and per-group letter choices. Compare the columns: Qāf is exact (57 = 3×19, no room to move), while a group like Nūn reaches 19² only once particular alif-forms are counted in. That gap is the fitting freedom the project's own audit measured.
The strata of the script.
One printed verse is really historical layers stacked: the consonantal skeleton (rasm) written first, and the vowel marks and notation added by later hands. Peel them back.
Printed canonical
Consonantal skeleton — vowel marks removed
Layer map — base letters vs added notation
The layer nobody documents.
The same verse, inside a computer, is not one thing. Its codepoints, its normalization, even its length depend on silent choices in every app and font. Here they are, exposed.
Three lengths of one verse
Same text, three honest answers — because "length" itself is a convention.
Normalization — NFC vs NFD
The alif family present
Codepoints
Read the Qurʾān.
Read any sūra straight through — the Arabic beside a translation in your language — and bookmark the āyāt you want to keep. Over a hundred translations across dozens of languages.
The readings, transmitted.
One consonantal text; several canonical ways of reciting it. Ten readings (qirāʾāt), each carried by two named transmitters (riwāyāt), were accepted as authentic. Almost every printed muṣḥaf today is one of them.
The ten canonical readings
Where the readings came from
A documented difference
The differences are real but bounded — vowelling, hamza, elongation (madd), a few consonant points, the basmalah between sūras. The most-cited single example:
Compare the readings for any verse
Six riwāyāt as real text, differences from Ḥafṣ highlighted. (Readings can number the verses differently — itself a documented ʿadad variation — so the same number can point to a shifted āya.)
The earliest witnesses.
Before print, the text was carried in parchment and ink. A handful of very early codices survive — the physical evidence behind every claim about how faithfully the text was transmitted. Below are real folios from the oldest and most important witnesses, each openly licensed and one click from its full record.
Search the whole text.
Every verse, searchable at once — in the Arabic or in any translation. A concordance for the entire Qurʾān: find where a word or an idea appears, then open any hit in Compare.
Method, and how to read the grades.
The atlas asserts documentation, not doctrine. Everything carries a provenance grade, a citation, and a version — so a claim made against one release can be checked against that release forever.
Provenance grades
canonical Grounded in verified data from quran.ai / quran.com — the canonical text, per-surah structure, and metadata. Reproducible against the source.
documented Reported in the classical literature and widely cited, with primary-source line-citation being added (e.g. the ʿadad totals, the collection reports, the reading differences).
seed Reserved for any claim not yet verified against a source. The catalog currently carries none — every item on display is canonical or documented.
How the atlas works
Every lens is computation or citation over grounded text — never assertion from memory. Where the scholars disagree, the disagreement is recorded as data; it is never resolved here. Numerical claims live in the Counting Lab as claims: reproduced faithfully, stress-tested against the freedom the documented variables allow, and never permitted to edit the text.
A quick glossary
The handful of terms worth knowing. Switch to Beginner in the top bar for plain-language explanations throughout.
Open, versioned, free
Code under MIT, the catalog under CC-BY. Each release is frozen and citable. Corrections and scholarly contributions — especially on the ʿadad tables, the qirāʾāt differences, and manuscript alignments — are the whole point. Get in touch →
Open API
Every lens is powered by a small, same-origin JSON API you can call directly — for research, replication, or building your own tools. All endpoints are GET, CORS-open, and edge-cached. No key required.
| Endpoint | What it returns | Example |
|---|---|---|
| /api/verse | One āya in five orthographies (Uthmānī, Imlāʾī, IndoPak, simple) | ?key=2:255 |
| /api/reading | Full verified text of a riwāya (list mode with no query) | ?riwaya=warsh |
| /api/counts | Per-sūra āya counts + totals across the six riwāyāt | /api/counts |
| /api/translation | 126 translations (list); a verse in chosen editions | ?key=2:255&ids=131 |
| /api/tafsir | 20 exegetical works (list); a verse in a chosen tafsir | ?key=2:255&id=169 |
| /api/words | Word-by-word Arabic, transliteration, and English gloss | ?key=2:255 |
| /api/search | Full-text concordance over Arabic + translations | ?q=mercy |
Underlying sources: Quran.com / Quran Foundation API, KFGQPC (King Fahd Complex) via jsDelivr, Tanzil. Please cite them, and the Atlas version, in published work — the Cite button in Compare generates a ready reference.
Sources & research.
A working bibliography and link index for everything the atlas draws on — the classical primary sources, the digital text projects, the manuscript archives, and the academic apparatus. A node you can start from.
Compare any verse
Pick any āya and set any two of its written forms against each other — mushaf Uthmānī, modern Imlāʾī, IndoPak, or the bare rasm skeleton. Same reading (Ḥafṣ); the words match, only the writing differs. Every difference is highlighted, character by character.
Tick a reading to add this verse in that transmission. Differences from Ḥafṣ are highlighted. (Readings can number verses differently, so the same number may point to a slightly shifted āya — flagged where it happens.)
Tafsīr is verse-by-verse scholarly commentary. Pick an edition and add it to the view; add more than one to compare how they read the same verse.
Every word of this verse, in order, with its transliteration and a short English gloss beneath. The Arabic is the Uthmānī script of the Ḥafṣ reading (the same verse shown above); it updates whenever you change the verse. This is a word-level study aid — for a flowing translation, use the Translation panel above.
Hear this exact verse recited. Choose a reciter and press play; the audio follows whichever verse you have selected above.