Why fancy text turns into boxes, and where it breaks

Those styled letters are separate Unicode characters, not styling. That explains every symptom.

Fancy text generators do not change the font. They swap each letter for a different Unicode character that happens to look styled — 𝓖 is U+1D4D6, a mathematical script capital A, not an A in a script font. If the device has no installed font containing that character, it renders the replacement box, nicknamed tofu. That single fact explains the boxes, the broken search, and the screen reader problem.

What is actually being substituted

Most fancy styles come from three corners of Unicode, none of them designed for decoration:

Because the letters come from different blocks, a single fancy string can be perfectly readable except for two characters that arrive as boxes. Small caps is the worst offender: the Phonetic Extensions block has no small capital q or x, so generators substitute something else or fall back to the ordinary letter, and the word looks wrong even when it renders.

Where it breaks, specifically

• Screen readers. VoiceOver and NVDA read 𝓖 as mathematical script capital A or skip it. A bio written in fancy text is announced as a stream of symbol names, so a blind visitor cannot read your business name.
• Search. Nothing matches. Instagram search, site search and Ctrl+F all compare code points, and 𝓖 is not A. Putting your brand in fancy text removes it from search.
• Older Android. The default font on many mid-range devices ships without the Mathematical Alphanumeric block. Boxes.
• Windows terminals, code editors, CSVs. Monospaced fonts almost never contain these blocks.
• Email subject lines. Some clients strip non-ASCII from subjects; several spam filters score unusual Unicode density.
• Forms and databases. A name entered in fancy text fails validation, or is stored and then never found again by anyone searching for the normal spelling.
• Copying back out. Paste a fancy name into a document and you cannot spellcheck, sort or case-convert it — a case converter has nothing to convert, because these characters have no case mapping.

Strikethrough and underline are different, and safer

Strikethrough text is not a substituted alphabet. It uses combining characters: after each normal letter, an invisible mark such as U+0336 combining long stroke overlay tells the renderer to draw a line through the preceding character. The letters stay real letters, so search still finds them and a screen reader still reads the word.

That makes strikethrough the one decorative effect that is reasonably safe in a bio or a chat message. Its failure mode is different: the stroke may sit slightly off centre, or double up on wide glyphs, and copying the text into a system that normalises Unicode can strip the combining marks and silently un-strike everything.

Character counts double, though. Each visible letter is two code points, so a 20-character strikethrough word costs 40 against an Instagram or Twitter limit. Check it in a character counter before you hit a hard limit.

What usually goes wrong

• It looks perfect on your phone. You tested on the device that has the font. Ask someone on a different platform to send you a screenshot before you commit a profile to it.
• Boxes appear only in some words. Coverage gaps inside a style. Switch style rather than trying to fix the letter.
• The name stops appearing in search. Predictable and usually discovered weeks later.
• Pasting into a CMS strips it. Many editors normalise input to NFKC, which converts mathematical alphanumerics straight back to plain ASCII. That is Unicode working as designed.
• Character limits hit early. Fullwidth and combining styles cost 2 or more per visible character.
• Mixed with invisible characters. Generators sometimes pad with zero-width joiners, which then break word count and search in the receiving system. If text behaves oddly after pasting, strip it with a text cleaner and start again.

Where to use it, where not to

There is a security angle too, and it is the reason some platforms strip these characters outright. Because the styled letters look like ordinary ones but compare as different code points, they can be used to slip past keyword filters, or to register a display name that looks identical to somebody else. Moderation systems have responded by normalising or rejecting the whole class of characters, which means a bio that renders today may be silently flattened after the next app update.

Reasonable: a single decorative word in a personal bio, a separator, a one-off comment, a nickname in a game chat. Anywhere the text is ornament and nothing depends on reading it.

Not reasonable: your business or account name, anything a customer needs to search for, headings on a web page, email subject lines, product names, form fields, alt text, and any text that must be read aloud.

If the goal is emphasis rather than decoration, real formatting always wins where it exists: bold in WhatsApp with asterisks, italic in Telegram, actual CSS on your own site. Where no formatting exists, line breaks and a bullet character do more for readability than a styled alphabet, and cost nothing in accessibility.

A middle path that survives everything: keep the account name plain and put the decoration in a line of the bio below it. Search still finds you, screen readers still read the name, and the styling lands where its failure costs nothing.

Related tools

See all text tools →

Frequently asked questions