Two Ways to Write Ai in Thai: ใ and ไ

Two shapes, one sound

ไป and ใจ rhyme perfectly. The first is bpai, to go; the second is jai, heart. Both end in the sound ai, and the letter carrying that sound is different in each one, for no reason you can hear.

The upside is that this part of the spelling is finite. Most spelling difficulties in Thai are open ended: the tone of a syllable depends on a class, a mark, a vowel length and a final consonant all at once, and new words keep arriving that need the whole calculation. This one does not. There are twenty base words spelled with ใ, and a word built out of one of them keeps its ใ, which is why การตัดสินใจ, a decision, is spelled with the rolled form. Every other ai takes ไ. The Thai Wikipedia article on ไม้ม้วน states both halves: new words all take ไ, except words derived from one that already has ใ.

The two letters have names that describe their shapes. ไ is ไม้มลาย, mai malai, split wood, for the way its upright divides. ใ is ไม้ม้วน, mai muan, rolled wood, for the curl at the top.

The twenty words that take ใ

Here they are. Read the glosses rather than trying to hold the shapes: several of these are words you will use on your first day.

Most of them are everyday words: ใน, ให้, ใช้, ใจ, ใคร, ใหม่, ใหญ่, ใกล้, ใส่ and ใช่ turn up in the simplest sentences. The two that look odd in a beginner list are สะใภ้ and ใหล. The second has no entry of its own in English Wiktionary at all, and survives only inside compounds such as หลงใหล, to be crazy about, and ใหลตาย, sudden arrhythmic death syndrome. Thai Wikipedia lists it as ใหล, followed by those compounds in brackets, and that base form is the one that counts as the twentieth.

Ten of these twenty are already sitting in Phayan, the app this blog belongs to, which teaches reading letter by letter from zero and carries audio on every word it introduces. ใจ, ใน, ให้, ใช้, ใช่, ใส่, ใส, ใต้, ใคร and ใหญ่ all have an entry with their tone and their spelling broken into parts, and the course introduces ไ and ใ together in a single round, opened with ไป and ใจ under the note: two spellings, one sound.

How many Thai words really carry ใ?

The rule says twenty, plus whatever is built from them, so it can be counted. On 26 October 2026 we took every title in English Wiktionary's Category:Thai lemmas: 17,379 lemmas, of which 933 are spelled with ไ and 452 with ใ. Nineteen of the twenty have a lemma of their own. Of the 433 other lemmas with ใ, 429 owe every ใ in them to one of the twenty, from การตัดสินใจ to ผู้ใหญ่, and not one of them is a loanword or a new coinage with a rolled vowel of its own.

The four that break the rule are the page for the letter ใ itself and three archaic entries: เชยีงใหม, an old spelling of Chiang Mai, ใชย, an old spelling of ชัย, and ใภ้, an archaic word for daughter in law. The table gives each of the twenty with the values from the dataset: whether Wiktionary has a lemma for it on 26 October 2026, how many other lemmas carry it, and whether Phayan's word list has an entry for it.

ใจ alone sits inside 168 other lemmas, from ใจดี to หายใจ. The table also caught a mistake in this post: until this check, its list of the twenty printed หลงใหล where the base form ใหล belongs.

The school verse that holds all twenty

Thai has taught the set as verse for centuries. According to Thai Wikipedia, a textbook of the Ayutthaya period, Chindamani, already listed the twenty in rhyme, and the kap yani verse it calls well known to this day strings them into a short nonsense story: ผู้ใหญ่หาผ้าใหม่ ให้สะใภ้ใช้คล้องคอ, an adult looks for new cloth for a daughter in law to wear round her neck.

The last line is the part worth knowing even if you never learn the rest: ยี่สิบม้วน จำจงดี, twenty rolled ones, remember them well. The verse counts itself. If you have twenty words and you can still recite the line, you have the whole rule.

Why the language kept both

Because they were not always the same sound. The rolled vowel is already in the Ram Khamhaeng inscription of the Sukhothai period, in ใคร, ใด, ใน, ให้, ใช้ and ใต้, and the Thai Wikipedia article on ไม้ม้วน describes its original value as a different diphthong, roughly a followed by the vowel of อึ, which standard Thai has since merged with ไ. Lao as spoken in Luang Prabang, Shan, Black Tai and White Tai still keep the old vowel apart, so a speaker of one of those hears a difference where a Bangkok speaker hears none.

That is the ordinary way writing systems end up with spare parts. English keeps the gh in night for the same reason: the spelling outlived the sound. Thai, for ai, kept exactly twenty.

Both are written before the consonant they follow

ไ and ใ belong to the group of Thai vowels that sit to the left of their consonant on the page while being pronounced after it. ใจ is written ใ then จ, and read jai: the consonant first, the vowel second. A reader coming from an alphabet that runs strictly left to right has to drop the habit of sounding out in written order here.

It is the same behaviour as เ, แ and โ, which is why the vowel forms are worth learning as a set rather than one at a time. Our guide to the thirty two written vowel forms lays out where each one sits around its consonant, and ไ and ใ are the last two of the four that do not pair up into a short and a long version.

Romanisation hides all of this

Write these words in the Latin alphabet and the whole distinction evaporates. The Royal Thai General System of Transcription romanises ใ and ไ alike as ai, so ใน comes out nai, exactly as a hypothetical ไน would, and ใส comes out sai, the same as ไส. Two spellings collapse into one, and the reader who learned the word from a phrasebook has nothing to fall back on when they meet it written.

This is the narrow version of a wide problem: romanisation is a pronunciation aid that throws away the information you need to read. Phayan measured how far that goes on Thai place names and published the dataset alongside it. The study on Thai place names in English has the numbers.

The practical upshot for anyone learning: do not try to learn which of the two a word takes from its romanisation, because the romanisation does not contain the answer. Learn the twenty, keep ใ in anything built from them, and let every other ai be ไ.

How was this checked?

On 26 October 2026 a Python script using only the standard library read every title in English Wiktionary's Category:Thai lemmas through the MediaWiki API, counted the lemmas spelled with ไ and with ใ, and tested every ใ against the twenty listed on Thai Wikipedia; the lemmas outside the twenty were then read by eye. The same script checks which of the twenty have an entry in Phayan's word list, and re-reads every page this post quotes. The results are in dataset.csv and the script is check.py, which exits 0 while every count and quote still holds.