Emoji in business texts: what they cost

Put one 🎉 on the end of a 147-character text and it goes from one part to three. Same words and the same phone, three times the price. The emoji didn't add a hundred characters. It changed which alphabet the whole message is sent in, and the alphabet is where the cost lives.

Most people who send texts for a living know that much, in the way you know a hob is hot. The detail is in two ETSI specifications and a few Unicode data files, and it's worth half an hour, because the rules decide what a text costs and they aren't quite what you'd guess. Some emoji take two of your places. Some take four. One common one takes eleven.

Where 160 and 70 come from

A text message carries 140 octets of user data. That figure is in 3GPP TS 23.040, the specification that defines SMS, and nothing about the handset, the network or the sending service changes it. What changes is how many characters fit into 140 bytes, and that's the job of the companion specification, 3GPP TS 23.038, which ETSI publishes as TS 123 038 and which defines the alphabets.

The one everyone means by "160 characters" is the GSM 7 bit default alphabet. Each character is seven bits, the bits are packed end to end with no gaps, and the standard does the sum itself: "in 140 octets, it is possible to pack (140x8)/7=160 characters". The table has 128 entries and they're an odd mix. All the English letters and digits, the usual punctuation, £ and $ and ¥, a run of accented letters (è, é, ù, ì, ò, à, ü, ö, ä, ñ), a set of Greek capitals for some reason, and § and ¿ and ¡. What's missing matters more. It has no curly quotes, no ellipsis character, no long dashes and no emoji.

The other alphabet is UCS2, which TS 23.038 covers in a line: the message "can consist of up to 140 octets, i.e. up to 70 UCS2 characters". Two bytes a character, so half as many characters. A phone or a gateway picks one alphabet for the whole message, and it has to pick UCS2 if even one character isn't in the 7-bit table. That's the cliff. There's no mixed mode where the emoji costs two bytes and the rest costs seven bits. One emoji, or one curly apostrophe pasted in from Word, and all 146 of your other characters are now two bytes each too.

Long messages make it worse. When a text runs past one part it's split into concatenated segments, and each segment gives up a few bytes to a header that tells the phone how to stitch the pieces back together. TS 23.040 gives the per-segment limits: 153 characters in the 7-bit alphabet ("160-7") and 67 in UCS2 ("(140-6)/2"). So the honest comparison for anything over one part is 153 against 67, and a 160-character message that needs UCS2 is three parts, not two. Our guide to the character limit goes through the arithmetic.

Which emoji take more than two places

The "70 characters" line undersells the problem, because an emoji is rarely one UCS2 character.

Sixteen bits reach only the first 65,536 code points. The Unicode Consortium's own FAQ describes the "remainder of about 1 million characters" as needing a pair of 16-bit units, called surrogates, and it's blunt about the older name: UCS-2 "cannot be used to conformantly represent supplementary characters", while the encoding phones use in practice, UTF-16, "sometimes requires two code units to represent a single character". Nearly every emoji people use lives above that line. The faces are in the Emoticons block, U+1F600 to U+1F64F, well past 65,536, so 😂 (U+1F602) takes two of your 70 places on its own.

Then it compounds. Unicode Technical Standard #51, the emoji specification, defines several ways to build one picture from several code points, and the phone sends every one of them.

The emoji presentation selector. Some symbols existed as plain text before emoji did, and get a trailing U+FE0F, "used to request an emoji presentation for an emoji character", so they render in colour. ❤️ is U+2764 plus U+FE0F: two places. The bare ❤ without the selector is one place, but keyboards hardly ever send it that way.

Skin tones. A modifier sequence is the base emoji followed by one of five modifier code points, U+1F3FB to U+1F3FF, and both halves sit above the line. 👍🏽 is U+1F44D U+1F3FD: four places.

Flags. A flag is "a sequence of two Regional Indicator characters", each above the line. 🇬🇧 is U+1F1EC U+1F1E7: four places.

Joined sequences. The zero width joiner, U+200D, glues people together. The family 👨‍👩‍👧‍👦 is man, joiner, woman, joiner, girl, joiner, boy: seven code points, four of them two units each, for eleven places. One emoji, eleven of your seventy.

The sequences in the chart come straight from Unicode's emoji-test.txt data file for Emoji 18.0, dated 30 April 2026, which lists every emoji with the exact code points a keyboard sends. If you want to know what a particular emoji costs, that file has the answer, and our SMS character counter does the sum for you.

What people send

Does any of this bite on the emoji people reach for? Unicode publishes the only frequency data that isn't a vendor survey, gathered to help it decide which new emoji to add, and its ranking groups emoji by how often they're used relative to the top one, with each group roughly half the frequency of the group above. The top group is two emoji: 😂 and ❤️. The next, 😍 and 🤣. Then 😊, 🙏, 💕, 😭 and 😘, and after that a group of eighteen headed by 👍, 😅 and 👏 that includes 🔥, 💪 and 👌. Every one of the nine in the top three groups is a two-place emoji, either because it sits above U+FFFF or, in the heart's case, because it carries the selector. The one-place exceptions are old text symbols that never needed the selector, ✅ and ⭐ among them, and they sit further down the list.

The standard keeps growing, too. Unicode's counts page puts the total at 3,972 emoji in Emoji 18.0. The pace has slowed from the 776 added in 2015 and 755 in 2016, and 2024 brought only 8, but 2025 added 163, nearly all of them skin-tone and family sequences, which are the expensive kind.

For a business the frequency data says something useful. The emoji your customers type (laughing, hearts, thumbs up, hands) are all two places, and the ones that cost four or more (skin tones, flags, families) are the ones that look most staged when a company uses them.

The characters that catch you without an emoji

Most Unicode texts I've seen a business send had no emoji in them at all. The culprit was a character that looks ordinary.

The 7-bit alphabet has straight quotes, ' and ", and nothing else. Type an apostrophe on an iPhone or paste a sentence from Word and you get a curly one, which isn't in the table, so the whole text drops to 70. The same goes for the ellipsis character autocorrect makes from three dots and the long dash it makes from two hyphens. None of these look odd on screen. They show up as a 2 next to the parts count.

There's a second trap inside the 7-bit alphabet itself. TS 23.038 has an extension table "reserved for symbols of international significance", reached through an escape character, and the standard says the displayable character count "will reduce by one for every instance" the extension is used. The characters on that table are the euro sign, square brackets, curly brackets, the backslash, the caret, the tilde and the vertical bar. Each costs two of the 160. A price written as €25 is one character to the eye and two to the network. Hardly a disaster, but a text that counts to exactly 160 with a couple of euro signs in it is two parts.

What it costs

Take a reminder:

Crown Dental: your check-up is on Thu 19 Nov at 10:30. Reply C to confirm.

That's 74 characters, all in the 7-bit table, so one part with room to spare. Add a space and a 😊 on the end and it's 77 UCS2 units, which is over 70, so it's two parts of 67. One smiley has doubled the price of a reminder that was half empty.

Now a marketing text:

Crown Dental: 20% off whitening for patients until Fri 28 Nov. Book on 01234 567890 or reply WHITE and we will call you back. Reply STOP to opt out

That's 147 characters, one part. Add a space and a 🎉 and it's 150 units, which at 67 a segment is three parts: 67, 67 and 16. CaptiFi Connect charges one credit a part, from 3.5p a credit at the largest pack. Three parts is three credits, so the party popper has trebled the price of every text in the campaign. Send that to 2,000 patients and it's 6,000 credits instead of 2,000, for a picture most of them will see for a second.

The same sum runs the other way, and it's the part people forget. If a text is under 70 units with the emoji in it, the emoji is free. A 44-character "Your table is booked for 8pm, see you then 🙂" is 45 units, so it's one part either way. The cost of an emoji depends entirely on where the text already sits against the two limits, and the only way to know is to count. Our SMS character counter shows the alphabet, the units and the parts as you type, and the glossary entry on Unicode SMS explains what the counter is doing.

When an emoji is worth it

An emoji earns its place when it does a job words can't do in the space. A ✅ at the front of a confirmation reads faster than "Confirmed:", and it's a one-place symbol, so it's cheap. A ⭐ in a review request says what the text is for before the reader gets to the verb. Both are old text symbols that happen to have emoji presentations, and both sit in the 2019 ranking, so people recognise them.

It doesn't earn its place as decoration on the end of a 147-character offer, and it rarely belongs in a service text at all. A reminder from a dentist or a message about a payment reads better plain; the inbox it lands in is full of other plain, serious messages, and a 🎉 next to the one from the hospital looks like it was sent by someone else. If you do want one, put it where it carries meaning, keep the whole text under 70 units, and check the count before you send, because the emoji the keyboard gives you is not always the one you think you typed. The thumbs up with a skin tone is four places, and if your keyboard has remembered a skin tone you'll get that version without noticing.