For Arabic it mostly doesn't work. The reasons are a useful tour of how Arabic is encoded, so here are the pieces, what each one does when it's pasted somewhere, and which are safe to use.

1. An Arabic letter is one code point, its shape is the font's job

Arabic is cursive. Most letters have up to four shapes: isolated, initial, medial and final. ب, for example, is drawn ﺏ alone, ﺑ at the start, ﺒ in the middle and ﺐ at the end.

Unicode stores one code point, U+0628 ARABIC LETTER BEH, and the text-shaping engine (HarfBuzz, CoreText, DirectWrite) picks the shape from the neighbouring letters. The style comes from the font: Kufic, Naskh, Nastaliq. That is why "copy the calligraphy" can't work. The style lives in the font, and a font doesn't travel with plain text.

2. The "styled Arabic" block exists, and it's a trap

Unicode does have Arabic Mathematical Alphabetic Symbols (U+1EE00–U+1EEFF), with initial, tailed, stretched, looped and double-struck letters. They look like exactly what a fancy-text generator needs. They're meant for mathematical notation, so:

• They don't join. Each one is a standalone symbol, so a word comes out as a row of separate glyphs instead of connected script.
• Few fonts cover them. On many phones they show as empty boxes.
• They aren't the letters. Search, find-in-page and screen readers don't treat them as Arabic words.

If you build a text tool, don't map Arabic input onto this block.

3. Presentation forms: legacy shapes and a few real gems

Arabic Presentation Forms-A and -B (U+FB50–U+FDFF, U+FE70–U+FEFF) encode the contextual shapes as separate characters, a leftover from systems that couldn't do shaping. You'll find them in old PDFs: copy text out of one and you often get U+FE91 ﺑ instead of U+0628.

Mostly they cause bugs. A string full of presentation forms only matches its normal spelling after NFKC normalization, and many search boxes, databases and spell-checkers don't normalize. If you handle user input, normalize it:

const clean = input.normalize("NFKC");   // ﺑﺴﻢ (presentation forms) → بسم (letters)

The exception, and the one thing people really can copy and paste, is the whole-word ligatures at the end of Presentation Forms-A:

Each is one character that draws a whole phrase in calligraphic form, and it pastes into almost any app. The caveats are real: the size and look depend on the app's font, and a few older devices show a box. Normalization is inconsistent too. NFKC expands ﷲ, ﷴ, ﷺ and ﷻ back into their letters, which is usually what you want for search, but ﷽ has no decomposition at all. It stays a single symbol that will never match the words بسم الله, so don't use it where the phrase needs to be findable:

"ﷺ".normalize("NFKC");   // "صلى الله عليه وسلم"
"﷽".normalize("NFKC");   // "﷽"  (unchanged)

Unicode 14 added more (U+FDCF, U+FDFE, U+FDFF). Font support for those is still thin, so test before you rely on them.

4. Tatweel: the only "style" that survives a paste

U+0640 ARABIC TATWEEL (ـ), also called kashida, is a stretching stroke placed between two joined letters. Calligraphers lengthen joins for rhythm and justification, and tatweel gives plain text a similar look:

بسم الله        →  بســـم الله
الحمد لله       →  الحمـــد لله

It is an ordinary character, so it survives any paste. Two rules:

1. Only between letters that join. ا د ذ ر ز و never connect to the next letter, so a tatweel placed after them floats on its own.
2. It breaks matching. بســـم is not بسم to a search index. For a username or a bio line it doesn't matter; for content people need to find, keep it out.

If you strip harakat in a tool, decide whether tatweel goes too. We strip it with the marks, because it isn't part of the spelling.

5. Joiners: showing a letter's shape out of context

To display a letter's medial form on its own, for example in a lesson that says "this is ب in the middle of a word", put a joiner on both sides:

const ZWJ = "\u200D";
const medialBeh = ZWJ + "ب" + ZWJ;   // displays the medial form ﺒ

U+200D ZERO WIDTH JOINER asks the shaper to join, and U+200C ZERO WIDTH NON-JOINER asks it to break a join. Use ZWNJ in Persian, where it is part of the spelling (می‌خواهم). In Arabic, ZWNJ is almost always an accident, so treat it as a bug in user input.

6. Harakat are combining marks, so mind the order

Short vowels and other marks (U+064B–U+0652, plus U+0670 dagger alif) are combining characters that stack on the previous letter. Two things to know:

• NFC reorders them. Shadda has combining class 33 and fatha has 30, so normalization puts fatha before shadda even if the user typed shadda first. Good fonts render both orders the same. If you compare strings with marks, normalize both sides first.
• Stripping them is a one-liner, which is useful for search and for logos:

const noHarakat = s.replace(/[\u064B-\u0652\u0670]/g, "");

7. Bidi: why your pasted bio looks scrambled

Paste @user مرحبا 2026! into a left-to-right field and the punctuation and numbers often land on the wrong side. Arabic is right-to-left; digits and neutral characters take their direction from what surrounds them.

The fixes, from least to most robust:

• U+200F RIGHT-TO-LEFT MARK (RLM): an invisible strong RTL character. Put it after trailing punctuation so the ! stays at the Arabic end.
• U+2067 RIGHT-TO-LEFT ISOLATE … U+2069 POP DIRECTIONAL ISOLATE: wraps a run so it can't disturb the text around it. Use it when you insert user names into a sentence.
• In HTML, use <bdi> and dir="auto" rather than inserting marks yourself.

So what should a "copy and paste" tool offer?

Our conclusion after building one:

1. Copy gives real Unicode letters, and you tell people plainly that the style won't come along.
2. An image or vector export (PNG, SVG with the font embedded) is how the calligraphic look travels.
3. A short list of the whole-word ligatures (﷽ ﷲ ﷺ), clearly labelled as symbols with their caveats.
4. Tatweel for people who just want a stylised bio line, with the joining rule spelled out.

That's what we built into our Arabic calligraphy generator. It has 22 open-licence fonts, it runs entirely in the browser, and it doesn't use any styled-letter trickery.

If the problem starts one step earlier, because you need to type Arabic on a computer with no Arabic layout installed, our online Arabic keyboard works in any browser. French speakers can use the same tool as a clavier arabe en ligne.

If you've hit an Arabic Unicode edge case that isn't covered here, tell me about it in the comments.