Fancy Text Compatibility: What Actually Works on Instagram, TikTok, Discord, WhatsApp, Facebook, X & Twitch

How these ratings were arrived at

Last reviewed 2026-08-20

Why a Box Appears, and Which Font Is Really Doing the Work

Measured 2026-09-28 against 10 font files

A box — the empty rectangle, sometimes called tofu — is not the platform refusing your text. It means the device looked through the fonts it has and none of them contained a drawing for that character, so it drew a placeholder instead. That part of the question does not need a lab: it can be answered by reading the fonts.

We read the character tables out of the fonts Android ships and the fonts installed on a Windows machine. Every style in the table above is covered by at least one font on both — so on a current Android phone or Windows machine none of them should be a box. Those are the two font sets that were read; an iPhone or a Mac may differ, and this check cannot say. The interesting part is which font covers them:

Not one of the lettering styles comes from the interface font. Bold, italic, gothic and monospace arrive from a maths font; bubble letters from a symbol font, which is a different file again. Two things follow, and they explain most of what people report. A styled word can look subtly mismatched beside ordinary text, because it is literally set in another typeface. And an app that pins one font to a field, rather than letting the system fall through to its fallbacks, has nowhere to find these characters — which is when the boxes appear even though the device owns a font that could draw them.

Underline and strikethrough behave differently again: they are ordinary letters plus a combining mark, so every font checked can draw them. Their failure mode is not a box but a misplaced line.

What this does not cover. iOS and macOS fonts are not available to this check, and nothing here describes what a platform does after rendering — stripping a character, rejecting a field, or normalising the text. Those remain reasoned, or checked by hand using the steps above.

What Three Browsers Actually Drew

Measured 2026-09-30 · Windows 11 Pro 10.0.26200, desktop

Reading a font file tells you a glyph exists. It does not tell you the browser will find it. So the next step was to put every letter of every style on a real page, in Google Chrome 154, Microsoft Edge 154 and Chromium 149, draw each one, and compare it against the box.

972 glyphs were drawn and 0 were boxes. A clean result is only worth reading if the test could have failed, so every run was also handed three Unicode code points that have never been assigned and therefore cannot be in any font. All three came back as boxes, in all three browsers. The test can see a box; there were none to see.

• All 52 letters of bold, italic, script, gothic, bubble and monospace drew a real glyph in all three browsers. 972 observations, zero boxes.
• The two combining-mark styles added no width of their own — the mark sits over the letter rather than beside it — and the mark was visibly drawn, not silently dropped.
• Styled text read back from a DOM selection code point for code point, unchanged, for all eight styles. That is the step a copy is built on.

The check most people write for this is wrong

There is a well-travelled trick for detecting a missing character: measure how wide it is, and compare that to the width of the box. It is quick, it needs no canvas, and it is the first thing this test did. It was also wrong, and worth showing why.

The box has one advance width — 64.5625px at a 100px font size in Chromium 149, Windows 11. Real letters are free to share it. Of 312 styled letters checked, 5 sit within half a pixel of the box while drawing perfectly:

Five of 312 styled letters measure within half a pixel of the box. A width-based check calls all five missing; all five draw correctly. If you are testing whether a character will show up, compare what is drawn, not how wide it is. The results above therefore come from comparing pixels: each letter is drawn into a canvas and its image compared with the box’s. Two characters that produce the same pixels are the same glyph, and no width coincidence survives that.

What this does not cover. One operating system, and three browsers that all use the same engine and the same installed fonts. Their agreement rules out a bug in one browser; it is not evidence about a different engine, and none about iOS, macOS or Android, which were not tested. Nothing here describes what a platform does to text after it renders — stripping it, rejecting a field, or normalising it.

Unicode Homoglyphs: Why Styled Text Is Not the Lookalike Problem

Measured 2026-10-01 · Chromium 149, Windows 11

A homoglyph is a character that reads as a different one. The word gets attached to styled text a great deal — 𝐕 is not V, so surely it is a homoglyph of V — and it is worth measuring rather than arguing about, because the answer changes what you should actually worry about.

Each character drawn at 100px, centred on the baseline, and compared with the plain ASCII letter it imitates. The figure is the share of inked pixels that differ: 0% means the two were drawn as the same shape.

Not one of 416 styled letters was drawn as the same shape as the plain letter it stands for. 36 of 39 cross-script lookalikes were. Styled alphabets are built to look different — that is the entire point of them — so they make poor disguises. The characters that genuinely deceive are ordinary letters from another script, and most people have never been shown one.

Eight of them, so you have seen one

Each pair below was drawn as the same pixels on the machine tested. The left character is not a Latin letter.

How this was kept honest. A letter compared with itself reads 0%, and letters nobody confuses — o against W, i against M, l against g — read 84% to 94% apart. A comparator that returned a plausible figure for everything would have measured nothing.

What this does not cover. One font stack, on one machine. A different font can pull a pair closer together or further apart, and a few of these shapes do differ in some typefaces — which is exactly why none of this should be relied on by eye. It is also a statement about shapes only: it says nothing about which of these characters a given platform will accept or reject.

This section answers whether the styles on this page are a risk. If what you actually need is to check something that was sent to you — a name, a link, a sender — that is a different job, and spotting a lookalike character covers it: all 36 pairs by code point, what your browser already does, and how to decode text rather than squint at it.

Which Style to Use: Work Down the List

Work down from the top. The first question that matches your case gives the answer; you do not need the ones below it.

1. 1. Does the text have to be typed back, searched for, or read aloud?

   Yes — Use plain letters. No style in this table survives any of those three — the characters are different code points, so search will not match them and a screen reader will not read them as words. This outranks every other consideration.

   No — Carry on to the next question.

2. 2. Is it going in a username, handle or @mention?

   Yes — Use plain letters. Not one platform in the table above permits styled characters in a handle, and the reason is the lookalike problem below, not fussiness. Style the display name instead, where the platform keeps them separate.

   No — Carry on.

3. 3. Will it be read on a device you cannot check?

   Yes — Pick from bold, italic, gothic or monospace, and keep the important words plain. Those four had two covering fonts on both font sets read here; script and bubble had one, so they have less to fall back on. Put the style on the decoration, not on the message.

   No — Carry on.

4. 4. Do you need an underline or a strikethrough specifically?

   Yes — Use the app’s own formatting if it has any — WhatsApp, Discord and Slack all do. The Unicode versions are a letter plus a combining mark, which is the one construction in this table that depends on the text engine rather than the font, and the only one whose failure is a misplaced line rather than a clean box.

   No — Carry on.

5. 5. Is it a short decorative line you can look at after posting?

   Yes — Use whichever style you like, then look at the posted result on a second device. Everything in this table drew correctly in the browsers tested; the variable is the reader’s device, and the only way to know is to look.

   No — If none of the above fits, default to plain letters with one styled word for emphasis. That is the choice that degrades most gracefully when something in the chain does not cooperate.