Paragraph Counter

Structure at a glance

The hero number is the paragraph count, but the figures beside it are what make it useful. Dividing sentences by paragraphs gives the average number of sentences per block, which is the quickest description of how a draft is built: a low ratio means short blocks, a high ratio means long ones, and a long paste that reports a single paragraph has not been divided at all. The 4-sentence sample worked through below runs 3 paragraphs, an average of 1.3 sentences per block. The ratio describes what you have; it is not a target, and this page does not prescribe a paragraph length.

Related guide: for how sentences, paragraphs, and lines relate to one another — and why the three counters on this site can disagree with the counts in other software — read How Sentences, Paragraphs and Lines Are Counted. The rest of this page covers the paragraph rule specifically, with every number computed by the engine that runs in your browser. More background lives in the guides.

What this counter calls a paragraph

The rule is the plain-text and Markdown convention: a paragraph is a block of text separated from its neighbours by at least one blank line. Mechanically, the engine splits the text wherever a line break is followed — after any amount of whitespace — by another line break, then discards any resulting piece that contains nothing but whitespace. Everything that survives is a paragraph. That one sentence explains every result on this page. A line break on its own is not a split point, because it is not followed by a second line break. A line of spaces between two breaks is still a split point, because whitespace is allowed between the breaks. A run of several blank lines is one split point rather than several, because the pieces between the individual breaks are empty and are discarded. It is a convention, not a law of writing — word processors use a different one, covered below — but it is the one plain text, Markdown, and web writing settled on, and it is the one a reader perceives when looking at a page of text.

Worked example: 3 blocks, one internal line break

Take this sample, shown with ↵ marking every line break:

The bridge opens at dawn.↵↵Traffic queues on the north bank.↵Nobody honks.↵↵By noon the river is quiet again.

The 2 blank lines divide it into blocks, and the second block contains a single line break between its sentences. The engine reports 3 paragraphs, 4 sentences, and 20 words. The internal break in the second block does not split it, because a break needs a second break after it — with only whitespace between — to count as a separator. That behaviour is deliberate: an address, a verse of poetry, or a list of short lines typed one under the other stays one block, which matches what a reader perceives and what Markdown would render. The line counter reports the same text as 6 lines — 4 non-empty, 2 blank — which is a different measurement of the same string, and the guide linked above explains why the two are worth keeping apart.

The same words with single line breaks

Now remove the blank lines and leave one line break between every sentence:

The bridge opens at dawn.↵Traffic queues on the north bank.↵Nobody honks.↵By noon the river is quiet again.

The word count is unchanged at 20 and the sentence count is unchanged at 4, but the paragraph count falls to 1: with no blank line anywhere, the whole text is one block. The line counter still sees 4 lines. This is the mechanism behind the copy-paste disagreement between this page and a word processor, and it is worth understanding rather than just noting. A document whose paragraphs were separated by paragraph marks alone, with the space between them supplied by formatting rather than by empty lines, can be exported to plain text with a single line break where each paragraph ended. Pasted here, that text looks exactly like the sample above — several visual lines, no blank line — and counts as one paragraph. The fix is mechanical: put a blank line between blocks, or check how the source application exports plain text, and the two counts converge.

Whitespace-only lines are blank

A separator line does not have to be empty; it has to contain nothing visible. In this sample the first separator line holds only spaces (shown as ␣) and the second holds only a tab (shown as ⇥):

First idea here.↵␣␣␣↵Second idea here.↵⇥↵Third idea here.

The engine reports 3 paragraphs from 3 sentences, exactly as if the separator lines were empty, because the split pattern allows any whitespace between the two line breaks. This matters more than it sounds. Editors that auto-indent can leave a line of spaces or a tab where you pressed Enter twice, some copy operations preserve trailing spaces on otherwise empty lines, and neither is visible on screen. Under the paragraph rule they behave as blank lines, so an "invisible" separator produces the same count as a visible one and a stray indent can never glue two paragraphs together. The line counter classifies the same lines as blank, so the two tools agree on what a blank line is.

Several blank lines in a row still make one break

Writers who space their blocks generously, or who paste from a document that used empty paragraphs for spacing, produce runs of blank lines:

Opening statement.↵↵↵↵Closing statement.

Here the engine reports 2 paragraphs, while the line counter reports 5 lines of which 3 are blank. The run of blank lines is one separator because the pieces between the individual breaks are empty and are discarded before counting. This is where the second, opposite disagreement with word processors comes from: they count every paragraph mark, including the empty ones used for spacing, so a document with generous spacing lines can report far more "paragraphs" than any reader would perceive. On the sample above, a counter that treated every line as a paragraph mark would be counting the 3 blank lines alongside the 2 blocks. The blank-line convention deliberately collapses them, and the paragraph figure it produces is the one the eye would give.

Leading and trailing blank lines, and empty text

The same discard step settles the edge cases. Blank lines before the first block or after the last one produce empty pieces, which are discarded, so this padded sample — ↵↵Only one block here.↵↵ — counts 1 paragraph, not more. And text that contains only spaces and line breaks produces nothing that survives the discard, so it counts 0 paragraphs, 0 words, and 0 sentences: the engine returns zeros across the board for empty or whitespace-only input rather than reporting a phantom block. Neither case affects an ordinary draft, but both matter if you are pasting fragments, and both confirm that only blocks with visible content are ever counted.

Reading the sentence and paragraph counts together

Sentences per paragraph is the most useful derived figure on this page, and it inherits the limits of both counts. The paragraph count is exact under its convention. The sentence count is not: as the methodology page documents, a sentence boundary is a run of . ! ? or … followed by whitespace or the end of the text, so abbreviations such as "Dr." or "e.g." each add a phantom sentence. The methodology's own worked paragraph shows the effect in miniature: "Dr. Smith arrived at 9 a.m. She left early." is 1 paragraph and 10 words, but 3 sentences rather than the human answer of two, so its sentences-per-paragraph figure comes out at 3. The bias is always upward, and on any draft longer than a paragraph it is diluted by the sentences that were counted correctly. Read the ratio as a description of block length that runs slightly high on abbreviation-heavy prose, not as a precise measurement, and use the sentence counter when the sentence figure itself is the question.

Where the copy-paste cases come from

Every disagreement between this page and another program reduces to one question: what separates the blocks after the text is converted to plain text? A word processor stores a paragraph mark at each press of Enter and counts those marks. This page never sees the marks — only the characters that survive the conversion to plain text — and it counts blank lines. When the same document is pasted here, the outcome depends on what that conversion produced. If each paragraph mark became a line break and nothing else, adjacent paragraphs arrive separated by single breaks and merge into one block, the case worked above, and the count here is lower. If the document used empty paragraphs for spacing, those become real blank lines here, the two conventions agree on the visible blocks, and the word processor's own figure is the one that runs high, because it counted the spacing marks. If the export produced a blank line between paragraphs, the two figures match. Which of these you land in depends on the source application and its export settings, not on this page — so when the number here surprises you, look at the pasted text in the box: if the blocks are touching with no empty line between them, that is the whole explanation, and adding blank lines resolves it. Text written directly in the box, in a plain-text editor, or in Markdown already follows the blank-line convention and counts the way it reads.

Check the rule yourself

The rule is small enough to test by hand. Type a sentence in the box, press Enter once, and type another: the paragraph count does not move. Press Enter a second time before the next sentence and it rises. Put a few spaces on the empty line between them and it does not fall back. Add extra empty lines and it does not rise further. Delete everything but the spaces and every figure returns to zero. Each observation corresponds to one clause of the split-and-discard rule described above, and the numbers in this page's worked examples are produced at build time by the same function that runs on each keystroke, so the prose and the tool cannot drift apart. The counting happens entirely in your browser: the page makes no network request while you type, and it keeps counting with the connection off.

Frequently asked questions

Counting happens locally in your browser — nothing you type is transmitted or stored. See the methodology page.