Hadith Explorer الكتب الستة

Method

Hadith Explorer is a free, open, searchable database of the hadith in the six canonical Sunni collections. Its aim is for every hadith to carry a quality level and a list of every concern raised about it, each with its source, whether the concern comes from Sunni hadith criticism, Shia hadith criticism or the historical-critical method.

Principles

  1. Transparency, not authority. There is no reviewer panel. Credibility comes from a published, fixed rubric applied the same way to every hadith, and from citing every concern.
  2. Every concern is cited. A concern is recorded only if it has a source: a named scholar's grade, a narrator assessment from a named biographical dictionary, or a published study. Grades and views are never inferred or paraphrased. Nothing is published as a concern until the source has been opened and the claim checked against it.
  3. Neutral, attributed language. The site reports views ("al-Albani graded it weak"); it does not issue verdicts or use polemical wording about any tradition.
  4. Gaps are shown. Each hadith shows, for each approach, whether it has been examined yet, so an unstudied hadith never looks clean by default.
  5. Responses are shown. Where a published response to a criticism exists, it is linked beside the criticism.
  6. Corrections are public. Every hadith page has a "report a problem" link, and fixes are listed in the changelog.

Three approaches

Sunni hadith criticism

Grades from classical and modern scholars, and narrator ratings such as Ibn Hajar's Taqrib al-Tahdhib. In this approach the contents of Sahih al-Bukhari and Sahih Muslim are regarded as authentic by consensus. The site shows that as a status of the collection, not as a grade of each report, and records classical critiques of specific reports (for example by al-Daraqutni) as concerns.

Shia hadith criticism

Narrator assessments from Shia rijal works (al-Najashi, al-Tusi, al-Khoei's Mu'jam Rijal al-Hadith) applied to identified chains, plus cited critiques of specific narrations.

Historical-critical method

Published academic studies: common-link and isnad-cum-matn dating, anachronism, textual development, and legal or political context. Examples include the work of Goldziher, Schacht, Juynboll (Encyclopedia of Canonical Hadith), Motzki, Schoeler, Görke and Little.

Concern records

Each concern records: the hadith, the approach, the concern type, its severity (minor, substantive or serious), a one-line neutral summary, the source citation with a link, any linked published response, and a note of what was checked in the source. A concern is published only after its source has been opened and the summary checked against it; candidates found by search or AI stay unpublished until then.

Concern types: weak or unknown narrator · broken chain · single chain (gharib) · hidden defect · narrator rejected in Shia assessment · conflict with reports from the Imams · conflict with the Quran or stronger hadith · wording varies across versions · dated later by common-link analysis · anachronism · legal or political context.

Methods named in concern records

Each concern says which method the cited study used, so readers can tell a narrator-based judgement from a dating study:

The site does not run these analyses itself or issue its own datings. It records the conclusions of named studies. From stage 3 it also shows each hadith's chain, so readers can check a study's claims against it.

Classical critiques of Bukhari and Muslim

Classical Sunni critics questioned a small number of reports in the two Sahihs, mostly on technical grounds in the chain ('ilal): a narrator added or left out, a disputed name, a report connected by one transmitter and left disconnected by others, or a phrase that may be a narrator's own words. The best-known collection is al-Daraqutni's al-Tatabbu'. Ibn Hajar's Hady al-Sari, the introduction to his commentary on al-Bukhari, lists in its eighth chapter the 110 hadith in al-Bukhari that al-Daraqutni and other critics faulted, and answers each one.

The site records each of these objections as a Sunni-lens concern and Ibn Hajar's answer as its published response. Each record quotes both, word for word, from the source texts. It applies only to the versions whose chain carries the criticised route, not to every version of the report.

Severity follows a fixed rule based on how Ibn Hajar, the defending scholar, treats the objection. No curator decides it:

Each record says who checked it. The first 94 were checked by an AI model (Claude) reading the source texts, with every quote confirmed by script. They have not yet been reviewed by a person, and each page says so. Corrections are welcome through the "report a problem" link.

Where it is unclear which of these applies, the curator's note on the record gives the reading taken and why.

Under the rubric, a minor concern moves a hadith in Bukhari or Muslim from High quality to Good quality. A substantive one moves it to Some issues.

What "examined" means

For each approach, a hadith page says either examined or not yet examined.

An examination that found nothing adds no concern. It records only that the gap was looked at, and it covers only the sources listed; it is not a finding that no critique exists elsewhere. Source sets are versioned, so adding a work creates a new version. The sets are listed on the Curation page.

Severity of Shia and historical-critical concerns

As with classical critiques, severity follows a fixed rule based on what the cited author says. No curator decides it.

Four quality levels

Each level is always shown with its name, never by colour alone.

The rubric

The rubric is code with tests (rubric.py, grades.py) and is applied identically to every hadith. Any change to it is listed in the changelog. The rules are checked in this order, and the first that matches sets the level:

  1. Major issues if any grader's grade is classed very weak or fabricated, or serious concerns are recorded in two or more lenses.
  2. Some issues if any grader's grade is classed weak, or any substantive or serious concern is recorded in any lens.
  3. High quality if the hadith is authentic with broad Sunni agreement (it is in Sahih al-Bukhari or Sahih Muslim, or at least two graders are recorded and every one is classed authentic) and no concern of any severity is recorded.
  4. Good quality if the Sunni lens supports it (at least one authentic or hasan grade, or it is in Bukhari or Muslim) but it falls short of High quality: some graders say hasan, only one grader is recorded, or only minor concerns are recorded.
  5. Not assigned otherwise: no classified grade, not in Bukhari or Muslim, and no concern that sets a level.

Cap (version 3): a report that gives the words or practice of a Companion (mawquf) or of a later figure (maqtu'), rather than the Prophet's, is at most Good quality, however strong its chain. The levels describe reports of the Prophet. See whose words a report gives.

Whose words a report gives

The cap needs to know which reports are not the Prophet's. Two kinds of evidence are used:

Limits: a Companion's own view that mentions the Prophet in passing is not caught, because reports that name him are taken to be his. Cross-references and unclear readings are not capped.

How grades are read

Grades are free text in the source edition. Each is read into one category, checking the most severe terms first, so a grade that mentions both a weak and a strong term counts as the weaker one:

Grades that fit none of these (for example "-", or "maqtu'" alone) are shown on the page but not used as a grade; a "maqtu'" label is used for the cap on reports not of the Prophet. Each hadith page lists every grade beside the category it was read as.

Provisional levels

A level is marked provisional while any of the three lenses has not yet examined the hadith. A lens counts as examined when at least one cited record from it applies to the hadith (for the Sunni lens: a grade, the Bukhari/Muslim consensus status, or a Sunni-lens concern). At present almost every level rests on the Sunni lens alone, so almost every level is provisional.

Modern scholarship counts

Concerns from the historical-critical method count in the level exactly like concerns from the other lenses: one substantive concern sets the level to Some issues, and serious concerns from two lenses set it to Major issues. Wherever such a concern exists, the hadith page shows it in a highlighted Modern scholarship box at the top, says what the level would be without it, and links any published response. All of them are listed on the Scholarship page.

Worked example

The hadith of ʿĀʾišah's marital age (for example Sahih al-Bukhari 5133). Sunni lens: in Sahih al-Bukhari, so authentic by consensus. Historical-critical lens: Little (2022) argues it was first put into circulation by Hisham ibn ʿUrwa in Kufah between 754 and 765 CE (substantive). Result: Some issues, provisional because the Shia lens has not examined it yet. Without the historical-critical concern it would be High quality. This case is one of the rubric's tests.

Chains of transmission and narrators

From stage 3 each hadith page shows its chain (isnad), split from the text by a program, with each narrator matched to an entry in Ibn Hajar's Taqrib al-Tahdhib. Where the same person appears in al-Tusi's Rijal, al-Najashi's Rijal or al-Tusi's Fihrist, that entry is quoted too.

Narrator dictionaries: OpenITI corpus (Ibn Hajar, Taqrib al-Tahdhib, JK000121; al-Najashi, Rijal, Shia002931; al-Tusi, Rijal, Shia002935, and al-Fihrist, Shia002934), licensed CC BY-NC-SA 4.0. Narrator data and pages derived from them are shared under the same licence.

What is in this release

Stage 4: the text of all six books, browsing, search, grades from the source edition, concern records, the rubric and the level of every hadith, each hadith's chain with automatic narrator matches, and the classical critiques of Sahih al-Bukhari listed in Ibn Hajar's Hady al-Sari. Also included: al-Daraqutni's and Ibn 'Ammar's critiques of Sahih Muslim with al-Nawawi's answers, al-Tirmidhi's own verdicts, and examinations of the 500 most widely transmitted reports in the Shia and historical-critical lenses. The Curation page shows the source sets, the progress and the studies not yet checked.

BookHadith shownWith per-hadith gradesEmpty source records set asideHigh qualityGood qualitySome issuesMajor issuesNot assigned
Sahih al-Bukhari7,580094,2691,7401,57100
Sahih Muslim7,36002035,1741,0621,12400
Sunan Abi Dawud5,2725,27221,7771,2442,195560
Jami' al-Tirmidhi3,9263,901721,1691,0051,6538118
Sunan al-Nasa'i5,6795,672862,1941,9561,493297
Sunan Ibn Majah4,3404,33831,3769061,7732832

"Empty source records" are entries in the source edition with no text in either language (for example, numbers reserved for sub-numbered reports). They are counted here and not shown as hadith pages.

Sources

Topics

Each book divides its hadith into chapters with its own titles. To browse one subject across all six books, each chapter title is mapped to a shared list of topics ("Prayer", "Trade and finance", "Faith and creed"...) by a fixed list of patterns, applied the same way to every book (topics.py). A topic says where the compiler placed a hadith; it is not a reading of the hadith itself. A hadith also carries the topics of the chapters where the compilers placed its other versions (see Versions), shown on its page as "versions also placed under". Al-Bukhari, for example, placed "actions are by intentions" under seven different chapters. The Browse page lists every topic and level, and each count opens a list of those hadith from all six books together.

Versions of the same report

The same report often appears several times: in different books, and within one book with a different chain or wording. Each hadith page lists the other versions found, and search can filter by how many books a report appears in. Versions are found by a script, applied the same way to every hadith (parallels.py, with tests). The groups have not been reviewed by a person.

  1. Text only. The text (matn) is compared, not the chain. Vowel marks are removed and letter forms normalised, and common words are set aside. Words that occur mostly in chains (narrator names, "he told us") are also removed, measured from all six books: where a chain was not fully separated from its text, it must not make two unrelated reports look alike.
  2. Score. How much of the shorter text is found in the longer, from 0 to 1: the average of the share of its word pairs found there, and the share of its rarer words found there, each word weighted by how rare it is. An abridged version inside a longer one scores high.
  3. Labels. A pair is same hadith when the score is at least 0.4 and the chains do not start from two different Companions. It is similar wording when the score is at least 0.3 but it falls short of that, including matching texts from different Companions, which in hadith terms corroborate each other (shahid) rather than being one report. A Companion is used only when Ibn Hajar's Taqrib places the narrator at the head of the chain in the first generation.
  4. Groups. Same-hadith pairs are joined into groups from the strongest down. Two groups join only when the average score over every pair between them is at least 0.3, so a single weak link cannot chain unrelated reports together.
  5. The compiler's cross-references. Some records have no text of their own and refer to the version before ("with this chain, the same", bi-mithlihi, bi-nahwihi), most often in Sahih Muslim. These join the group of the nearest earlier record in the same chapter that has a text, on the compiler's own statement.

Known limits: versions told in quite different words (for example first person against third person in a short report) can be missed. Where the Companion at the head of a chain is not identified, the same saying from two Companions can be grouped as one report. Concerns are not copied between versions: a concern applies to the versions whose chain carries the criticised route, as its record states. The 500-report curation queue keeps the grouping it was examined with (the first version of this script), so the examinations still cover exactly what was searched.

Links to the Qur'an

Each hadith page says whether the hadith explains a verse, quotes a verse, or has no link recorded, and each verse on the Qur'an pages lists the hadith linked to it. Links are found by fixed rules, applied by a script the same way to every hadith (quran.py, with tests). Nobody chooses links by hand.

A quotation is a run of consecutive words of a verse (or of consecutive verses) found in the hadith's Arabic, compared with vowel marks removed and letter forms normalised as in search. It counts when either:

Unmarked words that are not introduced as Qur'an are not counted. Many common formulas share wording with a verse, such as the dhikr "His is the dominion and His is the praise, and He has power over all things" (compare 64:1) or "whoever believes in Allah and the Last Day" (compare 65:2), and counting them would link hadith that are not quoting the Qur'an. Where the quoted words occur in more than one verse, every such verse is listed, and if they occur in more than six, no single verse is named. Each hadith page shows the matched words, so every quotation can be checked.

Explains a verse applies when any of these holds:

Quotes a verse applies to a hadith with at least one quotation that meets none of the conditions above.

No link recorded means none of these rules found a link. It is not a finding that the hadith has no bearing on the Qur'an. A paraphrase, an allusion, or a quotation shorter than 5 words that the editors did not mark is not caught. Al-Nasa'i's and Ibn Majah's Qur'an commentary is in other works, not in these collections. Later additions under consideration are a commentary that names the verse for each hadith (for example Ibn Kathir's) and a cited occasions-of-revelation work (al-Wahidi, Asbab al-Nuzul).

Numbering

Hadith numbering differs between editions (for example, Sahih Muslim is numbered both sequentially and by Muhammad Fu'ad 'Abd al-Baqi). Each hadith page shows the numbers the source edition gives: its own sequential number, the "Arabic number" field it supplies, and the chapter and position. Two numbers are never treated as the same hadith without a verified mapping; numbers from other editions will be added only with such a mapping.

Search runs in your browser using Pagefind; no query is sent to a server. Both the indexed text and your query are normalised the same way:

Corrections

If you find a mistake, use the "Report a problem" link on the hadith page, or open an issue. Please cite a source where you can. Accepted fixes are listed in the changelog.