G with Accent: How ğ, ĝ, ġ and ģ Turn a G into J, Gy or Gh

What sound does each g with accent stand for?

Almost never the g of gun. That is how the Fundamento de Esperanto (opens in a new tab) describes a plain g, and a mark on top is there to say the sound is something else. Which something depends on the writing system. Unicode's names list (opens in a new tab) names the users of most of these letters. The sounds are the IPA in Wiktionary's entries for ğ (opens in a new tab), ĝ (opens in a new tab), ġ (opens in a new tab), ģ (opens in a new tab), ǧ (opens in a new tab), ǵ (opens in a new tab) and ḡ (opens in a new tab).

Letter	Mark	Written in	Sound	Example
ğ	breve	Turkish	none of its own; the vowel before it grows long	dağ, mountain, /ˈdaː/
ğ	breve	Azerbaijani, Crimean Tatar	/ɣ/	dağ, mountain, [dɑɣ] in Azerbaijani
ĝ	circumflex	Esperanto	/d͡ʒ/, the j of join	manĝi, to eat
ġ	dot above	Maltese	/d͡ʒ/	ġobon, cheese
ġ	dot above	Irish in Gaelic type	what roman type spells gh	old print only
ģ	cedilla, drawn as a comma above	Latvian	/ɟ/	Ģirts, a man's name
ǵ	acute	Macedonian romanization	/ɟ/, for Cyrillic ѓ	place names in Latin letters
ǧ	caron	Skolt Sami	/ɟ͡ʝ/	a letter of its alphabet
ǧ	caron	Lakota, Dakota	/ʁ/	a letter of their alphabets
ǧ	caron	some Arabic transliterations	the Arabic letter ج, a j	romanized names
ḡ	macron	Arabic romanization	the Arabic letter غ	romanized names

Read the Sound column down and the marks fall into groups. Some push the g forward into a j: ĝ, Maltese ġ, and ǧ for Arabic. Some move it onto the hard palate, where /ɟ/ is made: Latvian ģ and ǵ. Some pull it back into the throat: Azerbaijani ğ, Lakota ǧ, and ġ and ḡ for Arabic. Turkish ğ goes furthest and drops out as a consonant. For the Latvian letter's odd shape, see the Latvian alphabet page.

Why can one g with accent be a j in one language and a gh in another?

Because no alphabet borrowed the meaning along with the mark. Each one picked a free mark for a sound it lacked, so the same g with accent can carry a j in one writing system and a throaty gh in the next.

Esperanto needed a j sound, since its own j is the y of yoke in the Fundamento. Its answer was ĝ, the g with circumflex. Maltese uses the dot for the softer of two sounds, Wiktionary notes, so ġ, the g with dot above, is the j and g stays hard. Irish used the same dot for a softened consonant, which is a gh.

The result is that one character can stand for opposite sounds. These readings come from the same Wiktionary entries as the table above.

Character	One reading	Another reading	A third
ĝ	Esperanto /d͡ʒ/	Aleut /ʁ/	Haida /ɢ/
ġ	Maltese /d͡ʒ/	Irish gh	Arabic غ in romanization
ǧ	Arabic ج in some transliterations	Lakota /ʁ/	Skolt Sami /ɟ͡ʝ/

Arabic shows the trap best. Its j letter, ج, is ǧ in some transliteration systems. Its throaty ghayn, غ, is ġ or ḡ in others. So a romanized Arabic name with ǧ and one with ġ hold different Arabic letters, though both look like a g with something on top.

Two habits follow. When you read a g with accent in a name, find out the language before you guess the sound. When you write one, copy the exact letter. A ǧ typed for ğ looks close, but it is a different character, and the caron page explains how to tell the two marks apart.

Is the Turkish g, ğ, silent in Azerbaijani too?

No. The letter is the same, U+011F, and the names list gives it to both Turkish and Azerbaijani. Only standard Turkish lost the sound. Wiktionary's Azerbaijani entry for ğ gives it the sound [ɣ], a g said without closing the throat, and its Crimean Tatar entry gives /ɣ/ as well.

The same spelling makes the difference audible. The words and transcriptions come from Wiktionary's entries for dağ (opens in a new tab), ağ (opens in a new tab) and Uğur (opens in a new tab).

Word	Turkish	Azerbaijani
dağ	mountain, said /ˈdaː/	mountain, said [dɑɣ]
ağ	net or web, said /ˈaː/	white, said [ɑɣ]
Uğur	a man's name, said [uːr]	not in the entry

So a Turkish speaker hears ğ as a longer vowel, while an Azerbaijani speaker hears a consonant.

The Turkish entry keeps a trace of the old sound. It lists [ɣ] for eastern dialects and says the breve marks a g that was dropped. It also notes that the Arabic letter غ turns into ğ inside a Turkish word, and into plain g at the start, because ğ never begins one. What ğ does after a vowel in standard Turkish is covered on the Turkish alphabet page.

What should you write when you cannot type ĝ, ġ or ğ?

What you write instead of a g with accent depends on the language, and only some languages have an official answer. Passports follow a rule of their own: the ICAO table turns Ĝ, Ğ, Ġ and Ģ into a plain G.

Esperanto has the clearest answer. The Fundamento says that when the marks cannot be printed, an h may replace the circumflex, so ĝ becomes gh. In 2007 the Akademio de Esperanto (opens in a new tab) allowed other signs as well, but only where the marks are impossible and gh does not suit, only as a technical means, and only where they cause no confusion. The best known is the x-system (opens in a new tab), which writes gx for ĝ.

Irish gh is not a substitute at all: it is how roman type writes the old ġ.

That table is the transliteration table in ICAO Doc 9303 (opens in a new tab), the standard for machine-readable travel documents. Ǧ, Ǵ and Ḡ have no row in it. For Macedonian, the UNGEGN report (opens in a new tab) gives two Latin forms of ѓ.

Letter and setting	Write instead	Who says so
ĝ in Esperanto text	gh	Fundamento
ĝ stored as plain ASCII	another code such as gx, as a technical means only	Akademio de Esperanto, 2007
ġ in Irish	gh, the normal modern spelling	Wiktionary
Macedonian ѓ, UN system	đ, or g before e and i	UNGEGN
Macedonian ѓ, orthography manual	ǵ	UNGEGN
Ĝ, Ğ, Ġ or Ģ in a passport's machine-readable line	G	ICAO Doc 9303
Ǧ, Ǵ or Ḡ in a passport	no entry in the table	ICAO Doc 9303

The gh stand-in has a weak spot of its own. Esperanto also has words where a real g meets a real h. Wiktionary's flughaveno (opens in a new tab), airport, is said /fluɡhaˈveno/, with a g and then an h. In text written with gh for ĝ, that same word reads as if it held a ĝ. Writers using gh have to watch for such joins. The Fundamento alphabet has no x, so gx cannot clash with a real pair of letters.

Where does each g with accent sort, and does a search for plain g find it?

Under the rules of a language that writes the letter, it is a letter of its own, and a plain g does not find it. We sorted and compared these words in Node.js v26.0.0, running ICU 78.3 with CLDR 48 data, on 2026-10-10. The search test used the strength that ignores case and accents. This is our own measurement, not a quoted rule.

The words come from Wiktionary: Maltese ġnien (opens in a new tab), garden, ġobon (opens in a new tab), gass (opens in a new tab), gas, and għajn (opens in a new tab), eye; Latvian gaiss (opens in a new tab), air, gulbis (opens in a new tab), swan, and the name Ģirts (opens in a new tab); Esperanto gazeto (opens in a new tab), newspaper, grava (opens in a new tab), important, and ĝardeno (opens in a new tab), garden.

Rules	Sorted by the language	Sorted by English rules	Plain spelling finds the letter?
Maltese, ġ	ġnien, ġobon, gass, għajn	gass, għajn, ġnien, ġobon	gobon misses ġobon
Latvian, ģ	gaiss, gulbis, Ģirts	gaiss, Ģirts, gulbis	Girts misses Ģirts
Esperanto, ĝ	gazeto, grava, ĝardeno	ĝardeno, gazeto, grava	gardeno misses ĝardeno
Turkish, ğ	Ufuk, Uğur, Uysal	the same	Ugur misses Uğur
Azerbaijani, ğ	af, ağ, ah	the same	ag misses ağ
English	not a letter there	not a letter there	every plain spelling matches

Maltese is the odd one: its dotted ġ comes before g, the way the CLDR Maltese chart (opens in a new tab) orders it. The Latvian (opens in a new tab) and Esperanto (opens in a new tab) charts put their letter after g, which is why Ģirts moves to the end. In the Turkish and Azerbaijani lists the order does not change: ğ comes after f and before h under both sets of rules.

The search column is narrower than it looks. In the same test, each language's rules kept apart only its own g letter. Maltese rules let a plain g match ğ, ģ and ĝ, and Turkish rules let it match ġ. If your users type without marks, an accent-blind search under English rules finds every one of them.

What do Uğur, Elazığ and ĝardeno look like after an accent strip or a bad decode?

Each g with accent is a g plus one combining mark, so the usual strip turns every one into a plain g. UAX #15 (opens in a new tab) defines the split, and we then dropped the marks in the same Node.js build. The second column is what happens when UTF-8 text is read as Windows-1252, decoded with the same build's TextDecoder. Uğur, Ģirts, Elazığ and Tekirdağ are from AccentPad's names and places facts.

Original	Marks stripped	UTF-8 read as Windows-1252
Uğur	Ugur	UÄŸur
Tekirdağ	Tekirdag	TekirdaÄŸ
Elazığ	Elazıg	Elazığ
Ģirts	Girts	Ä¢irts
ġobon	gobon	Ä¡obon
ĝardeno	gardeno	Ä, a hidden U+009D, then ardeno

Elazığ shows that a strip is not a full fix for Turkish. The ğ loses its breve, but the dotless ı is a letter with no mark to drop, so the result still holds a non-ASCII character. A slug or a file name made this way needs an explicit ı-to-i rule.

The ĝardeno row is the hardest to spot. In UTF-8, ĝ is the bytes C4 9D. Windows-1252 shows C4 as Ä, and the cp1252 mapping file (opens in a new tab) leaves 9D undefined, so the decoder passes it on as an invisible control character. On screen the word reads as Äardeno, with nothing to show that a byte is hiding there. Reopen the file as UTF-8 rather than retyping the letters.