WordBench

What changed, and when

Anything that changes what you see, what a number means, or what this site promises about your text is listed here with a date. Platform limits were last rechecked against their sources on 23 July 2026.

Sort Lines is now Alphabetize a List, with four new options

Changed

Most people who open this tool want to alphabetize a list, so the page now says so; the address /tools/sort-lines stays the same. Ignore articles files The Beatles under B, Sort by last name files Jane Doe under D, Case sensitive puts Apple before apple, and Remove empty lines drops blank lines first. Numbers inside a line are still read by value, so item 2 comes before item 10, and a line under the result says what was done.

Remove line breaks, with or without paragraphs

Changed

The extra spaces tool is now Remove Line Breaks and Extra Spaces, because joining lines broken by a PDF or an email is what most people come for. Join lines has two settings: Keep paragraphs joins the lines inside each paragraph and keeps one blank line between them, Join everything puts the whole text on one line. A line under the result counts the line breaks and extra spaces removed, and the page answers how to clean text from a Word document.

Upside down text: four modes and the full character table

Added

Next to Upside down there are now Mirror, Reverse words and Flip letters only. Every flipped character is now a single Unicode character; the old capital B was two. Characters without a convincing lookalike are left as typed and listed above the result instead of being faked, and the page shows the complete table with the code point of every character and its result, plus the keyboard characters that won't flip. Both tables are generated from the code the tool runs.

Small caps: the alphabet with code points

Added

The small caps page now has a table of every letter the generator changes, with the code point of the letter and of its small capital, generated from the code the tool runs. The letters that do not become a small capital are named on the page and listed above the result whenever they appear in your text.

Word frequency: phrases, a sortable table and CSV download

Added

The word frequency counter now also counts 2-word and 3-word phrases, never across the end of a sentence or a line, can skip common words, sorts by count or alphabetically, and downloads the full table as a CSV file that Excel opens with accents intact. The definition of a phrase is on the methodology page.

Fits everywhere: where to cut, and a copy that fits

Added

When text runs over a limit on the word counter or the character counter, the row for that field now shows where the text has to end and offers Copy trimmed to fit. The cut is measured in the unit that field counts in, ends on a whole word and never splits an emoji. The character counter also gained a table of common lengths, 150 to 2,200 characters, with the fields that use each one and roughly how many words it holds.

A short answer at the top of every tool page

Added

Every tool page now opens with a 40 to 70 word answer that says what the tool does and how its options work, so the answer is there before the scrolling starts. Each one describes the tool as it actually behaves, without figures that are not on the page.

Night paper, a dark theme

Added

WordBench now follows your device's dark mode, with a warm dark palette rather than black, and a button in the header switches between light and dark. The choice is remembered in this browser and applied before the page is drawn, so there is no flash of the wrong theme. Every text and background pair passes a 4.5:1 contrast check in both themes, and an automated check fails if one stops doing so.

Export, print and named drafts on the word counter

Added

One Export menu downloads your text as .txt or .md, the counts as a CSV file, or prints the text with its counts on a clean page. Drafts keeps up to 20 named texts, each with an optional word or character goal and a words-today count, stored in this browser's own database and nowhere else; the privacy page says what a draft keeps and how to delete it, because Clear in the editor leaves saved drafts alone. Opening a file that is not text, or one over 2 MB, now gets a clear message, and the counter keeps working.

The character limits as open data

Added

The limit table is now also published as /limits.json and /limits.csv, one row per field with the platform, the limit, the unit, the source and the date it was verified, licensed CC BY 4.0. Both files are built from the same data as the limit pages, so they can never disagree with them.

Limit pages answer the question in their title

Changed

Each limit page title now carries the number itself, for example the Instagram bio limit and the caption limit, and the page opens with one sentence stating the limit, the unit, the source and the date it was checked. More of the ways people ask the question are answered in the FAQ. No limit changed: the figures are the ones verified on 23 July 2026.

Two guides with clearer answers and bigger tables

Changed

How many words is 500 characters now leads with the answer, about 88 to 100 words with spaces, answers the average and with-spaces questions directly, and its table runs from 100 to 5,000 characters, every cell calculated. The words per page guide now has a table, using only figures the guide already sources or calculates.

Popular jobs on the home page, and a search box on the 404 page

Added

Under the counter on the home page, plain links go straight to the jobs people come for most: alphabetizing a list, removing line breaks, flipping text, small caps and the Instagram bio limit. A page that does not exist now offers a search over every tool, plus the tools you used recently, instead of a dead end.

Use as input, and clearer empty and loading states

Changed

Every tool with a result now has Use as input, which puts the result back in the text box so the next step can run on it, with one step of Undo. An empty tool explains what to paste and offers a sample, a tool that is still loading holds its place instead of making the page jump, and on a phone Copy and Use as input stay within reach at the bottom of the screen.

Methodology: sorting, line joining, phrases, trim to fit and character tables

Changed

The methodology page now writes out the rules behind this round's features: natural sort order and how articles and last names are compared, how Keep paragraphs decides where a paragraph ends, what counts as a 2-word or 3-word phrase, how the trimmed-to-fit copy is found in each field's own unit, and where the upside down and small caps tables come from.

Four statements that did not match what the tools do

Fixed

The small caps page said capitals you type keep their full height; they never did, every letter becomes a small capital. The word frequency example said the appears 4 times in a sentence where it appears 5. The reverse text page said every emoji stays intact; an emoji built from several code points, such as one with a skin tone, does not. And the methodology page said closing the tab clears your text, while each tool keeps it in this browser until you press Clear. All four now say what actually happens.

Character limit pages, one per platform

Added

Every platform WordBench tracks a limit for now has its own page at /limits, with the published number, the unit that platform actually counts in, what happens when you go over, a link to the source document and the date it was last checked. The data behind them was already on the site; it now has somewhere to live.

Methodology page

Added

Every counting rule the site applies is written out at /methodology: what counts as a word, how characters are measured four ways, how sentences are detected, the readability formulas, and the complete stop-word list used for keyword density. A density figure calculated against a word list nobody can see is not a reproducible number, so the list is published in full.

Embeddable word counter

Added

The counter can now be embedded on any site with one line of HTML, free, with no tracking and no cookies. The only condition is that the credit link stays visible and followable.

Readability scores on the word counter

Added

Flesch Reading Ease and Flesch-Kincaid Grade Level, with both formulas and the syllable heuristic published on the methodology page. Dale-Chall was deliberately left out: it needs a 3,000-word list, and shipping a guessed version would produce a number that looks authoritative and is not.

Headings that read as broken English

Fixed

Section headings were built by pasting the tool name into "About the ...", which produced "About the slugify" and "About the repeat text". Every tool now carries its own written-out heading, and the build fails if one is missing or reads as "the" followed by a verb.

Three meta descriptions that were cut off in search results

Fixed

Three descriptions ran past the length Google will show, so they were truncated mid-sentence in the result. The build now checks every title and description for length, duplication and a proper ending sentence.

Sitemaps split by page type, with real dates

Changed

One flat sitemap stamped every URL with the same date, which tells a crawler nothing. There are now four sitemaps behind an index, one per page type, and each URL carries the date its own content last changed.

AI crawlers are explicitly welcome

Changed

robots.txt now names eighteen AI crawlers and allows all of them, and the obsolete Host directive is gone. The reasoning: there is no paywall, no advertising revenue an AI answer could take away, and being named as the source is the only distribution this site has.

Platform character limits verified

Data checked

Every limit rechecked against the platform's own documentation. The unit column was added at the same time, because X counts weighted units, Bluesky counts graphemes with a second ceiling in bytes, and SMS counts septets or 16-bit code units depending on a single character in the message.

SMS character limit guide

Added

The 160-character SMS is the most misquoted limit on the web. It is only true for GSM-7; one emoji or one curly quote drops it to 70 for the whole message.

How this is kept up

Platform limits change without announcement, and a stale number is worse than no number. Every limit on the limit pages is rechecked against its source document each quarter, and every benchmark figure on the methodology page is rechecked each year. Both dates are shown on the pages themselves, so you can see how fresh a figure is before you rely on it.