Most styles in a text generator replace one character with another. Unicode underline and strikethrough do something completely different: they leave your letters alone and stack an invisible mark on top of each one.
That difference is not a technicality. It changes what the resulting text can do, and it makes these two the only styles that keep your words findable while still looking different.
Two invisible characters
U+0332 is COMBINING LOW LINE. U+0336 is COMBINING LONG STROKE OVERLAY.
Neither has a shape of its own. Each attaches to the character immediately before it and draws a line through or under that character. The text stays exactly what it was, and the marks sit on top of it.
So an underlined a is not a special underlined character. It is the letter a followed by U+0332, two code points that a renderer combines into one visual unit. Delete the mark and you have an ordinary a back.
Why that is useful
Because the letters underneath are unchanged, combining marks keep three things that substitution destroys.
Searchability. A search for the plain word still matches the underlined version in most systems, because the letters are identical. Some implementations compare exact byte sequences and will miss it, but the word is genuinely there to be found, which is more than can be said for a substituted phrase. The general problem is covered in does Unicode text hurt your SEO.
Coverage. Substitution styles only cover the characters Unicode encoded, which means 52 Latin letters and, for some styles, 10 digits. Combining marks work on every character: digits, punctuation, accented letters, Cyrillic, Greek, Arabic, anything. There is no alphabet they do not reach.
Screen reader behaviour. A reader that ignores combining marks announces the plain word correctly, which is far better than reciting mathematical character names or skipping the run. Some readers do announce the marks, which adds noise rather than losing the sentence. That is a much better failure mode, and it is covered in what screen readers really announce.
How the two approaches compare
| Property | Substituted letters | Combining marks |
|---|---|---|
| Underlying text | Replaced | Unchanged |
| Searchable | No | Usually |
| Works on digits | Only some styles | Yes |
| Works on other scripts | No | Yes |
| Screen reader | Symbol names or silence | Usually the plain word |
| Character cost | One per character | Two per character |
| Rendering consistency | Depends on font coverage | Depends on font mark positioning |
| Spell check | Fails | Usually works |
The last row is quietly important. Because the letters are real letters, a spell checker still reads them, so you do not lose autocorrect the way you do with substituted text.
Why it sometimes looks wrong
The mark is drawn by the font, so quality varies more than people expect.
On some systems the underline sits too low and reads as a separate line rather than an underline. On others it does not extend the full width of the letter, leaving visible gaps between characters so the line looks dashed. Strikethrough sometimes sits above or below the optical centre of the letter.
None of that is fixable from your side. It is the font’s decision about where to place a combining mark, and different fonts on different systems decide differently. The same variability that affects substituted text applies here for a different reason, as covered in why italic text looks different on every device.
The trailing space problem
A combining mark attaches to whatever character precedes it, including a space.
If a converter applies the mark to every character including trailing whitespace, you get a stray line hanging off the end of the phrase with nothing under it. This is the single most common visible flaw in underlined text, and it is the fastest way to spot a careless tool.
The fix is to trim the phrase before converting it, or to delete the final mark manually. A careful converter skips whitespace, which is what the Italic Text Generator does.
Doubling the length
Each character is now two code points instead of one.
On a platform with a strict character limit, an underlined message costs twice as much as it appears to. Fifty visible characters consume a hundred. On X that matters and can be the difference between a post fitting and not. In a document or a chat app with a generous limit it does not.
There is a second consequence. Some systems count characters by code points and some by grapheme clusters, so an underlined phrase can be accepted by one interface and rejected by another as too long. If you are near a limit, test in the tool you actually publish with.
Where they get stripped
Combining marks are more fragile in transit than substituted characters, which is the one place substitution wins.
Some paste paths drop a trailing combining mark, particularly on iOS, so the last letter of an underlined phrase loses its line. Some form fields normalise input using NFC or NFKC, which can reorder or remove marks. Some databases strip characters they classify as non spacing.
The practical workaround is to paste the phrase into a plain notes app first, confirm it survived, and copy it from there into the destination. If the line vanishes in a specific field, that field is normalising and no amount of retrying will change it.
Which platforms need them
Underline is the gap. Almost no social platform offers a native underline, which is why combining marks get reached for more often than any other decorative trick.
| Platform | Native underline | Native strikethrough | Combining marks needed |
|---|---|---|---|
| Discord | Yes, __text__ |
Yes, ~~text~~ |
No |
| No | Yes, ~text~ |
For underline only | |
| Telegram | Yes, in the menu | Yes, in the menu | No |
| No | Yes, ~~text~~ |
For underline only | |
| YouTube | No | Yes, -text- |
For underline only |
| Facebook and Instagram | No | No | For both |
| X | No | No | For both |
| No | No | For both | |
| Messenger | No | Yes, ~text~ |
For underline only |
The pattern is worth noticing. Platforms with a real formatting layer mostly cover strikethrough because it has an editorial use: showing a correction, a struck price, a changed plan. Underline is rarer because in digital typography an underline conventionally means a link, and platforms avoid a style that could be mistaken for one.
That convention is also a reason to think twice before using it. An underlined phrase in a caption reads as clickable to a lot of people, and they will tap it. If the underline is decorative, expect that confusion.
Strikethrough has the opposite problem: it is understood instantly, which makes it one of the most useful styles in this entire category. A struck price next to a new one, or a struck sentence followed by a correction, communicates something no other style can.
When to reach for them
Use combining marks when you want the text to still function as text.
A term you want people to find. A price you want compared. A correction that should still be readable, which is what strikethrough is actually for: showing what changed rather than hiding it. A heading that has to remain searchable.
Use substitution styles when the shape is the point and the words are decoration: a display name, a sign off, a two word flourish where nobody needs to search or hear the letters.
Stacking more than one mark
Nothing stops you putting both marks on the same character. A letter followed by U+0332 and U+0336 is underlined and struck through at once, and renderers generally cope.
Generally is doing work in that sentence. Mark positioning is a font decision, and two marks on one base character is a case fonts test less thoroughly than one. The common failure is the two lines sitting at different vertical offsets from letter to letter, so the phrase looks uneven rather than doubly styled.
There is also a limit to how far this goes. Long chains of combining marks on a single character are what produces the deliberately broken looking text people paste as a joke, and several platforms now cap the number of marks they will render per character or strip the excess. Two is safe. Ten is a platform’s problem to solve, and different platforms solve it differently.
If you want both effects, apply them and check the result on a phone before publishing. If the lines look ragged, pick one.
Try it
- Underline a phrase and paste it into a plain text editor.
- Use the editor’s find function to search for the plain spelling. It should be found.
- Now do the same with an italic substituted version of the same phrase. It will not be found.
- Underline a phrase containing digits and an accented letter, and confirm both got their line.
- Paste the underlined phrase into a chat app and check whether the trailing mark survived.
Steps two and three take ten seconds and demonstrate the entire difference between the two mechanisms.
Questions people ask
How does Unicode underline work? U+0332 is inserted after each character. It has no shape and draws a line under the character before it, leaving the letter itself unchanged.
Is underlined text still searchable? Usually, because the letters are unaltered. Some exact match implementations will miss it, but the word is genuinely present.
Why does my underline have gaps? The font decides how wide to draw the mark. Some fonts do not extend it the full width of the letter.
Why is there a line after the last word? A trailing space received its own mark. Trim the phrase before converting.
Do they count against a character limit? Yes, two code points per character, so roughly double the apparent cost.
The distinction worth keeping
Substituted letters change what your text is. Combining marks change how it looks while leaving what it is intact. Almost every practical difference between the styles in a text generator follows from that one line, and once you can tell which mechanism a style uses, you can predict how it will behave in search, in a screen reader and in a character counter without testing it.
