Bible Search

Which search should I use?

All of it lives at Bibles → Search, under two tabs: Keywords and Pattern (regex). Start here if you are not sure which one you want.

• I know the reference — type it in the Keywords box. he6.10, Heb 6:10 and Hebreos 6:10 all work. It is recognised as a citation automatically.
• I remember some of the words — type them in the Keywords box. Longer words tolerate a typo; short ones are matched as typed.
• I want two ideas near each other — still the Keywords box. Type both words, then widen the Words together chips that appear above it. See Two ideas, a few verses apart.
• I only want to search inside one passage — put the reference in front of the words: romans 3 faith, or romans 3-8 faith for a run of chapters.
• I want either of two words, or want to exclude one — type OR or NOT in capitals. See Either word, or neither.
• I want to know how often a word is used — the Concordance, which counts rather than lists. See The concordance.
• I want every form of a word at once — Pattern tab.
• I want the original Hebrew or Greek word, not one translation's wording — Pattern tab, then the Strong’s toggle.
• I am not sure of the word — just start typing in the Keywords box. From three characters it answers as you go and completes with words your own Bibles actually contain. See Results while you type.
• I can describe what I want but not write it — use the Ask box at the top of the Pattern tab and let it write the query for you.
• I only want part of the Bible — tap a preset above the search box, or just type gospels shepherd. See Searching part of the Bible.

Looking up a reference

Type a citation into the Keywords box and it is detected as a reference rather than as words to search for. You get the passage itself, shown from every translation you have installed, one card each.

• Full names — Ezekiel 34:25, 26
• Abbreviations — Phil. 4:6, 8-10
• Compact keys — eze34.25,26
• Prefix shortcuts — he6.10 resolves to Hebrews
• Several at once — Heb 6:10; Heb 11:6
• A whole chapter — ge1

Spanish book names are recognised too, so Hebreos 6:10 and Salmo 23 work without switching anything. Where a translation stores paragraph breaks, the passage is laid out in flowing paragraphs with superscript verse numbers rather than one verse per line.

A reference with words after it searches inside it

A citation on its own is a lookup. A citation with other words is a search inside that citation. Type romans 3 faith and you do not get the whole of Romans 3 — you get the verses in Romans 3 that contain faith, and the Searching: line above the box says Romans 3 so you can see the narrowing that was applied.

• A chapter — romans 3 faith
• A run of chapters — romans 3-8 faith. This is worth knowing: a chapter range is something only this form understands. As a plain lookup, Romans 3-8 resolves to Romans 3 alone.
• A verse range — romans 3:21-31 righteousness

An out-of-range reference is rejected rather than quietly trimmed, so Romans 99 is treated as words to search for rather than as a chapter that does not exist.

Keyword search

Anything that is not a citation is searched as words, across every installed translation at once. Results are ranked by relevance, with your search terms highlighted, and each one links through to the verse in context.

• Spelling is forgiving on longer words, by exactly one letter. A word of seven letters or more still matches with one letter missing, added or changed — rightous finds righteous, forgivness finds forgiveness. Two mistakes is one too many, and that includes swapping a pair: recieve does not reach receive, because moving the i past the e counts as two changes rather than one. Shorter words are matched as typed, because one wrong letter in a short word usually spells a different real word: faith would otherwise return every saith. Partial words match as prefixes at any length — right will find righteousness.
• All your words must appear. Several words narrow the search rather than widening it.
• Quote it for an exact phrase. "everlasting life" returns only verses containing that literal run of words.
• Filter by translation. Chips above the results show a count per translation; tap one to see only that Bible, tap All to come back.
• The list stops at 200. A broad word says Showing 200 of 1,347 results — the count is the true total, only the list is cut. The strip underneath is drawn from the whole set too, so narrowing to a book or widening the words is how you see the rest, not scrolling.

Either word, or neither: OR and NOT

Several words normally narrow a search — all of them have to appear. Three operators change that, and they are typed rather than clicked:

• OR — either word. shepherd OR pastor finds verses using one or the other, which is the search to reach for when translations disagree about a word.
• NOT — exclude a word. faith NOT works finds verses about faith that do not mention works.
• AND — the default, written out. faith AND works is the same search as faith works; it exists so you can be explicit inside brackets.
• Brackets group. (shepherd OR pastor) NOT hireling applies the exclusion to both alternatives rather than only the last.
• Spanish operators work too — O, Y and NO mean the same things.

Type faith or hope in lowercase and it searches for three words, exactly as it always has. The page notices, and offers you a chip: Capitalise OR to search for either word. Taking it changes only the operator, not the rest of what you typed.

When a boolean query cannot be answered

An empty result and an unanswerable question look identical — both give you no verses — so the page distinguishes them rather than letting you read a mistake as a finding about scripture:

• Nothing to search for. NOT faith on its own would match nearly every verse in the Bible, so you are asked to add a word rather than handed thirty thousand results.
• An operator with nothing after it. faith OR says so: OR needs a word after it.
• An unclosed bracket is named as such rather than silently ignored.

Results while you type

From the third character onwards, the results list, the count and the canon strip all update as you type. Nothing is submitted; you are just being shown the answer as it takes shape. A small live badge sits where the timing usually goes, so you can always tell a live answer from a committed one.

This is for the search you cannot quite phrase. You know there is something about faith and works in Paul; you are not sure whether the word is lovingkindness or loving-kindness, neighbour or neighbor. Typing used to cost a full round trip per guess and told you nothing about the guess itself. Now a word that is not in your translation stops producing results while you are still typing it, which is the moment that fact is useful.

Three characters, and why

Below three characters a prefix matches most of the Bible and stops being an answer — one letter matches around eleven thousand verses in the ASV, three matches under a thousand. So previews start at three.

Deleting back below three does not wipe what is on screen. The results stay, and the caption keeps naming the query they actually came from rather than what is currently in the box. That is deliberate: removing a letter should not destroy the answer you were reading, and results labelled with a query that did not produce them is the one mistake you could not catch by looking.

Why the count can go up when you press Enter

Live results match more strictly than committed ones, in two ways:

• No forgiving spelling. A half-typed word is a prefix, not a typo. Treating it as a typo offers you matches that vanish the moment you finish spelling it correctly — and turns faith into saith along the way. The forgiving match comes back on Enter, where it belongs.
• Only the last word is treated as unfinished. Type faith wor and faith is searched as the complete word you have finished typing, while wor is completed. Treating both as prefixes would silently search faithful and faithless too, which is neither what you asked for nor what Enter will do.

So faith wor might show 40 live and 57 after Enter. The number is not drifting — it is honest about which search ran.

What it says it searched

Live search only uses the translations already loaded, which may be fewer than you have installed. When they differ, it says so:

A count drawn from two of your four Bibles is a different measurement from one drawn from all four, and you could not tell them apart unless told. When the line is absent, live search covered everything you have.

Words the Bible actually uses

As you type, a list drops down of real words from your own translations, each with the number of verses it appears in:

Every row is a word drawn from the index itself, so choosing one cannot lead to an empty result. That is the whole difference between this and ordinary autocomplete. And because your English and Spanish Bibles share one pool, a bilingual library completes in both at once — which is the point, not a glitch.

• Ordered by how common the word is, not by how well it scores. Relevance ranking put shephuphan and shepho — one verse each, out of the genealogies — above shepherd, which is in ninety-six. Rare words score well precisely because they are rare, which is the opposite of useful when you are completing a word.
• The counts are exact, and count the word you would actually search. shepher reports two, not ninety-six: Mount Shepher, and not every shepherd.
• Very common prefixes get no completions. Counting each candidate means searching for it, and for a stop word like the that is a noticeable freeze in return for completing a word nobody searches on purpose.

Seeing a reference before you commit

Type a citation and the verse it resolves to appears beneath the box, before Enter:

A citation and a keyword search look identical while you are typing them and behave completely differently on submit — one returns a passage, the other a list of results. This makes that fork visible while you are still choosing it, instead of as a surprise afterwards.

It shows one verse from one translation, and counts the rest without reading it: John 3 shows verse 1 and offers Press Enter for all 36 verses. Unlike live keyword results this works from cold — it reads the chapter text rather than a search index, so it needs no earlier search to warm anything.

Where live results do not apply

• Pattern and Strong’s stay on submit. A half-typed pattern is usually an invalid one — typing [a-z]+ passes through four consecutive syntax errors — and a box that spent most of its life flashing red at someone doing nothing wrong would be worse than no preview.
• Widened Words together searches stay on submit too. Grouping a half-typed term into passages either empties the screen or regroups everything on every keystroke.
• A reference with words after it — romans 3 faith — commits rather than previewing.

Two ideas, a few verses apart

Type two or more words and a row of chips appears above the search box, labelled Words together:

In the same verse is the default, and it is what every multi-word search has always done. The other three are the interesting ones, because a verse is a citation unit rather than a unit of meaning. Chapter and verse divisions were added in the 16th century for referencing; an argument routinely runs three or four verses. Searching two ideas within a verse is close to searching within a page number — it works, and the boundary has nothing to do with the thought.

The case to try is any two words you expect to find argued together rather than stated together — faith and works, grace and law. Search them plainly and you get only the verses that happen to contain both. Widen to within 5 verses and you get the places where one leads to the other, which is usually the passage you were actually looking for.

Each result is headed with its reference, the terms it satisfied, and how far apart they were — Romans 3:21-28 · faith + works · 7 verses apart. That last figure is the tightness signal: two ideas three verses apart are far more likely to be one argument than the same two thirty verses apart.

What changes when you widen

• You get passages, not verses. Results render as the continuous span from the first matched verse to the last, with your terms highlighted in place. The verses between the matches are the argument connecting the two ideas — they are the reason you widened, so they are shown.
• Every term must still appear. Widening the window does not loosen the requirement. A passage containing one of your words twice is not an answer to “where do these two ideas appear together”, and it is not returned as one.
• Matching becomes exact. This is the one that surprises people. Ordinary keyword search is forgiving about spelling, which is right for find me verses about hope and ruinous here: one letter turns faith into saith, and a search for faith near works fills up with thus saith Jehovah. Naming two specific words near each other means those words. Prefix matching stays, so faith still reaches faithful.
• Tightest first. Passages are ordered by how close the terms are, so same-verse results stay at the top. Widening adds results below what you already had rather than reshuffling them.
• Windows never cross a chapter. The last verse of one chapter and the first of the next are adjacent on the page and are not one thought.

The chips only appear when there are two or more words to relate, and the setting travels in the link alongside the query and the scope, so a shared search arrives with its window intact.

Searching part of the Bible

Above the search box is a line that always tells you where you are searching. By default it reads Searching: Whole Bible. Narrow it and it says so — Searching: Gospels, or Searching: Romans 3 if you narrowed by typing a reference.

That line never collapses into an icon or a count of filters, and that is deliberate. A narrowing you have forgotten about is how an empty result quietly becomes a false belief about Scripture: you search a word, find nothing, and conclude it is not there. “3 filters” does not prevent that. “Searching: Gospels” does.

Three ways to narrow, and only one to learn

• Tap a preset — Gospels, Pauline Letters, All Prophets, Law, Hebrew Scriptures, Greek Scriptures. Any of the 66 books is under All books.
• Just type it. Type gospels shepherd and you are offered a chip: Search shepherd in Gospels. No syntax, no model call, no connection needed.
• Share the link. A scoped search carries its scope in the URL, so sending someone the link sends them the same search.

Where these verses are

Under the results is a strip showing how your matches fall across the Bible, grouped into the nine kinds of writing it contains. It is worth reading even when you have not narrowed anything, because it answers a question the result list cannot: where does this idea actually live?

Toggle it between Count and Per 1,000 verses. That second reading is the interesting one. Raw counts measure how long a book is as much as they measure your search — the Minor Prophets are barely 3% of the text, so a word can be far more characteristic of them than of the Law and still show a shorter bar. Searching shepherd makes the point:

Read that last phrase carefully, because it is doing real work. The comparison is against the books that actually use the word, not against the whole Bible. Most of Scripture contains no given word at all, and measuring against all that silence would make almost any book look remarkable — the sentence would fire constantly and mean nothing. Comparing like with like means you only see it when something is genuinely concentrated, and when there is nothing to compare against, you get no sentence rather than a flattering one.

The strip is also how you navigate

It is not only a readout. Tapping a bar does two things at once: it narrows the search to that kind of writing, and it opens the books inside it with their own counts. Tap a book from there and the search narrows again to that book alone.

This is usually the fastest way to work. Search a word across the whole Bible, look at where it clusters, then tap straight into the cluster — no need to decide in advance which part of Scripture you wanted. To come back out, use Clear on the Searching: line.

When a narrowed search finds nothing

This is the screen worth knowing about. Instead of a bare “no results”, a scoped search that comes up empty tells you where the word actually is:

Search everywhere clears the narrowing completely and returns you to the whole Bible. It does not half-widen.

This appears only when a narrowed search found nothing. It is the empty-result screen, not a panel that sits there on every search telling you what is elsewhere.

What it applies to

Narrowing works for keyword, exact-phrase and pattern searches, and it survives switching between them. It does not apply to Strong’s search, which uses a different engine — so the bar hides itself in Strong’s mode rather than sitting there looking as though it were doing something.

The box remembers your last twelve searches

Click into an empty Keywords box and a panel drops down with your recent searches, newest first, each with the number of results it found. Picking one restores the whole search — the words, the part of the Bible you had narrowed to, and the Words together window — not just the text.

The hit count is recorded after the search actually ran, so it is what was found rather than what was hoped for. A row reading 0 is worth seeing: it tells you that you have asked this before and it came up empty, which saves you asking twice.

Driving it from the keyboard

• ↓ / ↑ — move through the list. Pressing ↓ on a closed panel opens it on the first entry.
• Home / End — jump to the first or last entry.
• Enter on an empty box — runs the most recent search. The panel is already open and its first row is usually the one you want.
• Enter with something typed — searches what you typed. A remembered search has to be chosen deliberately with the arrows before it wins over the words in the box.
• Esc — closes the panel if it is open; empties the box if it is not. One key for “never mind”, at both depths.

The caret stays in the box the whole time, so arrowing down and changing your mind costs nothing — just carry on typing.

Clearing, and what survives it

The × inside the box — or Esc with the panel closed — empties the query and puts the recents panel back in front of you. Without it the panel was effectively unreachable: it opens only on an empty box, and after a search the box holds your query with no way to remove it.

Your scope and your Words together setting deliberately survive. This clears the query, not the search settings — someone narrowed to the Gospels trying three words in turn should not have to re-apply the narrowing each time. The Searching: line has its own Clear for the other intent.

Individual rows can be dropped with the × at the end of the row, and Clear in the panel header forgets all of them. Recents live in this browser only; they are never uploaded.

The concordance

Bibles → Concordance answers a different question from search: not where is this word but how often is it, and do my translations agree about it? Type a word and you get a row per installed translation.

• Occurrences, the way a printed concordance counts them. A verse using the word twice counts twice. This matters more than it sounds: in the ASV Jehovah is in 5,822 verses and appears 6,889 times, so nearly a fifth of the count lives in verses that say it more than once. Strong's and every printed concordance report the larger number, and so does this.
• The verse count is shown too, because that is the number the search page reports and the two should never look like a contradiction.
• A translation with no index is not counted, never zero. Zero would say the word is absent from a Bible nobody actually looked in — the one claim a concordance must not make by accident. Open that translation once to build its index.
• The canon strip appears here too, so you can see where the word lives while you are counting it, and See the verses hands the word to the search page.

Seeing what changed between translations

Open the comparison panel — the verse number in a chapter, or the verse page — and where you have two translations in the same language there is a Differences toggle. It marks the words that actually differ, so you are not left doing the comparison by eye.

• Same language only. Diffing English against Spanish would mark every word, which looks exactly like a finding and means nothing. The panel groups by language first and only offers the toggle where a real comparison exists.
• Punctuation and capitals are ignored. God, and God are the same word wearing different typography; marking those would bury the differences that matter. The text you see is always the original, accents and punctuation intact — the normalising is only used for matching.
• One English Bible and one Spanish is the default install, which gives nothing to compare. If that is you, the panel says so and points at the catalog rather than hiding the feature.

Pattern search

The Pattern (regex) tab searches by shape instead of by word. It is the tool for questions an ordinary search box cannot express: every form of a word at once, two ideas within a few words of each other, or a structural property such as verses that are questions.

Options

• Ignore case — on by default. Turn it off when capitalisation is the point.
• Ignore accents — off by default. Turn it on to make oracion match oración, which saves a lot of typing in Spanish.

Two rules the pattern box enforces

A verse is the unit of a match, so ^ and $ mean the start and end of a verse, not of the whole Bible. Patterns that could match across a verse boundary are rejected with an explanation rather than quietly returning a match belonging to neither verse — that includes the s flag, \n, [^] and [\s\S].

A search stops at 5,000 matches and says so. If you see that notice, the pattern is broader than you meant.

Strong’s search — the original languages

In the Pattern tab, switch the toggle from Pattern to Strong’s and type a number: H430, G26. Loose forms are understood too — h430, G 26, even Strong's H430.

This is a different question from every other search on the page. A keyword or pattern search asks which verses use this English wording, and the answer changes with every translation. Strong’s asks which verses use this Hebrew or Greek word — and that answer is the same in all of them, because it is a fact about the original text rather than about anyone's rendering of it.

• It works in every translation you have, including ones no other tool can search this way — the New World Translation, Reina-Valera, any Spanish or untagged Bible. It never looks at the translation's words, so it does not need them tagged.
• No setup, no storage cost. Unlike pattern search, Strong’s needs no opt-in and builds no per-translation index.
• A gloss header explains the word — the code, the lemma in Hebrew or Greek, its transliteration, and a short English meaning.
• A bare number finds all its senses. Searching H430 covers the disambiguated forms the data splits it into, listed under Also used for.
• No AI is involved at any point. You type the number yourself; there is no assistant path into Strong’s search. The verse list comes from published original-language data and a lookup on your device, so there is nothing here that could be hallucinated even in principle.

Details worth knowing

• Spanish glosses cover Greek only. The Greek source carries a Spanish column and the Hebrew source does not, so New Testament codes show a Spanish meaning and Old Testament codes do not. That is a gap in the data, not in your Bible.
• Psalm superscriptions are real results. They appear as Psalm 3 (superscription), never as “Psalm 3:0”. Most translations fold the superscription into verse 1, so the reference may be listed with a note that your Bible stores no separate text for it.
• Common words are enormous, and the count tells you so. H430 reports something like 2,603 matches in 2,248 verses and then renders the first 300. The counts describe the whole result set — only the list is capped, never the number.
• Attribution stays with the results. The original-language data is STEP Bible, used under CC BY 4.0.

Ask in plain language

At the top of the Pattern tab is an Ask box. Describe what you are looking for in ordinary words — every form of the word justify, verses that end in a question — and it writes the query for you, into the visible box, where you can read and edit it before or after it runs.

It also tells you, in one sentence, what the query it wrote will match, and lists the words it decided to search for when it widened a concept. That sentence is the point: a query can be subtly wrong in a way its results do not reveal, and reading the explanation is how you catch it without reading the pattern.

When it hands the query back instead of running it

Not every question is a pattern question. Ask decides which kind of search actually fits, and for many questions the honest answer is a plain keyword search — ranked and forgiving, which a pattern is not. When it decides that, it writes the query, explains it, and stops, telling you to run it from the Keywords tab.

That is deliberate, and worth knowing rather than reading as a failure. The quickest way for a feature like this to end up worse than the plain search box it was meant to help with is to reach for a pattern on every question. Being told “this is best run as a keyword search” means it picked the better engine, not that it gave up.

Ask always writes into the Pattern box, and switches you back there if you were on the Strong’s toggle — the toggle names the engine that actually ran, so it is never left pointing at one that did not.

Ask needs a connection. Offline, the box is disabled and you can still write a pattern yourself. If the assistant is unsure, it says so above the query rather than presenting a guess as an answer.