What this policy commits us to
- Character data is verified against the Unicode Standard, never copied from another symbol site.
- No symbol on a library page is published until it has been pasted into a real phone.
- On those pages characters are stored byte for byte. We do not tidy, normalise or substitute one.
- Visitor submissions are the exception, and this page says exactly how they are treated instead.
- Corrections are made on the page itself and the updated date moves with them. We do not backdate.
- No invented authors, no invented numbers, no paid placements.
This page describes how COOLMOJI decides what to publish and what has to happen before a symbol reaches a page. It exists because in this niche the failure modes are specific and mostly invisible: a character that looks fine in a preview and arrives as an empty box on a phone, a code point copied wrong from another site, a combination quietly broken by being “cleaned up” before it ever reached a TikTok display name or an Instagram bio.
Last updated: 12 August 2026 · Published by the COOLMOJI Editorial Team
Why we refuse second-hand character data
We do not take character data from other symbol sites, ever. It is the fastest way to publish and the fastest way to inherit somebody else’s mistake.
A wrong character name is unusually durable. One site publishes it, others copy it because it is already “on the internet”, and years later it is repeated everywhere without a single person having checked it against the standard that defines it. Nobody notices, because the only way to notice is to look it up — which is precisely the step that was skipped. Where our data comes from is set out on the about page; the rule here is what we do with it.
Kaomoji and decorative combinations sit outside that rule, because they are compositions rather than standardised characters. There is no register to check them against, so they are judged on one thing only: whether they survive being pasted somewhere real.
How we test that a symbol works
A symbol is drawn by the font on the device reading it, which means a preview inside our own editor proves nothing at all. Nothing on a library page is published on the strength of how it looks on our screen.
The about page lists the five checks and what each one catches. What it does not say is the order they run in, or what happens when one fails — which is the part that actually shapes a page.
The Android check runs first, not last
It is the first check rather than a final sanity pass, because a device-font gap is the most common failure and the cheapest one to catch early. Finding it first changes what gets written: the symbol usually stays, and the page gains a line telling you it depends on the device.
The iPhone check exists to stop us generalising
Apple ships a different font stack, so the same clipboard content can be correct on Android and meaningfully different on iOS. When the two disagree, the page has to describe both outcomes rather than pick the prettier one and present it as the result you will get.
The social bio check catches what no device test can
Platforms are stricter than operating systems, and their rules are invisible until you hit them. A failure here is not a font problem and cannot be fixed by choosing a better character — so it becomes a warning on the page instead, which is how the note about TikTok usernames came to be on this site at all.
The unicode.org check is the one aimed at us
The other four test the symbol. This one tests the editor. Names and code points are verified against the standard rather than against memory, because “I am fairly sure that is what it is called” is exactly how a wrong name enters circulation.
The last check happens after publication, not before
Copy buttons are tested on the live page rather than on a draft, because a grid item can look right in every respect and still put the wrong characters on your clipboard. A draft cannot prove that; only the published page can.
Every page carries the date it was last checked, in the line under its heading, so you can judge how fresh a claim is instead of trusting it indefinitely. That date moves only when the page genuinely changed, which is what makes it worth reading at all.
What we never do to a character
On our library pages, characters are stored and served exactly as they arrived, byte for byte. It is the most important rule we hold ourselves to, and the one with the least visible consequences when it is broken.
Concretely, we do not strip variation selectors, we do not remove combining marks, we do not convert a special or styled character back to plain ASCII, and we do not replace a curated combination with a more common emoji because it looked odd in a preview. A combination that appears broken in an editor very often pastes perfectly into a social app, so “it looks wrong to me” is never grounds for changing it.
Five real cases from our own pages. Every character below is reproduced byte for byte, and each one is a case where an automatic “clean-up” would have destroyed the thing people came to copy.
Swipe the table sideways to read the last column.
| Kept as stored | What a clean-up would do | Why it stays |
|---|---|---|
| ♡︎ · ♥︎ · ☠︎︎ | Strip the invisible variation selector sitting after each one | That selector is what forces the flat text style. Remove it and the character turns into a colour emoji — a different look from the one that was chosen |
| 💀ྀིྀིྀིྀིྀི | Delete the stacked combining marks as corruption | The stack is the effect. It looks like an error in a text editor and renders as intended in a social bio |
| ꧁⎝ 𓆩༺☠︎︎༻𓆪 ⎠꧂ | Replace the unfamiliar brackets with ordinary ones | Javanese, Egyptian and Tibetan marks in one string. Parts of it box on some phones — that is a missing font on that device, not a damaged string |
| ¯\_(ᐛ )_/¯ | Trim the “stray” space and escape the backslash | The space is part of the face and the backslash is an arm. Tidy either one and the kaomoji stops being a kaomoji |
| ⋆. 𐙚 ̊ · ᥫ᭡ · 𑣲 · 𖹭 | Swap them for common emoji that display everywhere | These sit outside the ranges older phones ship fonts for. They are exactly what people search for, and they paste correctly into TikTok and Instagram |
The consequence is deliberate: what you copy today keeps working tomorrow, because nothing is silently rewritten underneath you.
The one place that rule does not apply
Nicknames submitted by visitors go through a different path, and it does change the text. We would rather set that out than let the paragraph above imply a guarantee it does not cover.
A submission is cut to 30 characters, which can sever a long decorative combination part-way through. Runs of consecutive spaces are collapsed into one, so a combination that relies on double spacing for its shape will not survive intact. Entries built from private-use code points — the ranges no two devices agree on — are filtered out of the display entirely rather than shown and left to break.
None of that happens to the curated sets on our library pages, and none of it is done to make something look tidier. It is the cost of accepting text from the public, and the honest thing is to name it here rather than in a sentence nobody reads.
How we decide what gets a page
By measurement where measurement exists. The generator records how often each name is built and how often a generated result is copied, and those counts are the strongest signal we have about what people are actually looking for.
Two blocks on the home page are live readings of exactly that data, and it is worth being precise about which is which: Top Week Emojis Copied lists whatever was copied most inside the generator over the last seven days — results it produced and nicknames other visitors contributed alike — and the recent-names block lists the names built most over the last four weeks. Neither is a count of taps on the library grids — copying there happens entirely in your browser and is not recorded at all, which the privacy policy sets out in full.
We do not create a page simply to have another page. If a topic can only be covered by rephrasing what already exists elsewhere, it does not get published — a page has to add a tested example, a comparison or a correction that is not already available.
How corrections work
Corrections are made to the page itself and the updated date moves with them. There is no change log burying the fix somewhere nobody reads.
We do not touch the date to make a page look fresher than it is. The updated date moves when the content genuinely changed, and not otherwise — a page that has not needed a correction keeps its original date, which is information in itself. Anyone can report an error through the contact page, and reports that name the device and the app are the ones we can act on fastest.
Who writes these pages
Everything here carries the COOLMOJI Editorial Team byline, and that is a policy rather than a placeholder. We do not manufacture individual author profiles, portraits or job titles to make a page look more authoritative than it is.
The rule for changing that is simple: a personal byline appears only when a real, named person wrote the page and is willing to stand behind it. Until someone meets that test the team name is the honest attribution, and no page here will ever carry an author who does not exist.
What we will not publish
- Invented people. No fictional authors, photographs or credentials, and no borrowed expertise.
- Paid placements. We do not sell links, sponsored posts or guest articles, so nothing on the site is here because someone paid for it.
- Claims we cannot support. Where a symbol’s behaviour depends on the reader’s device or on a platform’s rules, the page says so rather than promising a result we do not control.
- A page built only from other pages. If we cannot add a tested example, our own comparison or a correction, we do not publish it at all.
Two more rules sit alongside those, set out with the rest of our limits on the about page: we do not invent numbers, and we do not reproduce copyrighted character artwork.
When this policy changes
If the way we work changes, this page changes with it and the date at the top moves. The commitments above are meant to be checkable against the site itself: if you find a page that breaks one of them, that is a mistake rather than a policy, and telling us about it is the fastest way to get it fixed.
Frequently asked questions
How does COOLMOJI check that a symbol is correct?
Character names and code points are verified against the Unicode Standard, and the symbol is pasted into a real Android phone, a real iPhone and a social bio field before a library page is published. The copy buttons are then tested on the live page.
Why does COOLMOJI keep symbols that look broken?
Because a symbol that renders as an empty box on one device very often pastes perfectly on another, and into apps like TikTok or Instagram. Characters on our library pages are stored byte for byte and never normalised, so removing one for looking odd would break it for the readers it works for.
Are visitor-submitted nicknames tested too?
No. Submissions are filtered automatically — length-limited, spacing-normalised and screened for blocked terms — but they are not pasted into a device by an editor before they appear. That difference is described in full above.
Who decides which emoji sets get a page?
Measurement from the generator does most of it: which names are built and which generated results are copied. Taps on the library grids are not recorded, so they play no part in the decision.
Does COOLMOJI accept paid articles or sponsored links?
No. Nothing on the site is published because someone paid for it, and pitches for links, sponsored posts or guest articles are not accepted.
Does COOLMOJI use AI to write pages?
Pages are drafted with software assistance and edited by the team. What matters more than the tool is the standard the page is held to, which is the whole of this policy: claims verified against the Unicode Standard, symbols pasted into a real device, no invented figures and no invented authors. If you find a page that falls short of that, it is a mistake rather than the standard — tell us and it gets fixed.