AI in Croatian: Diacritics, Address and Replies

AI in Croatian has reached the point where an assistant can write a clear reply, summarise a long email or explain a legal term in fluent Croatian. It still makes small mistakes that a native reader notices at once: a dropped diacritic, a wrong case ending, a word borrowed from a neighbouring language or a tone that is too casual for a customer. Knowing where those mistakes come from makes them easy to catch.

This guide explains how to use AI in Croatian for business and everyday writing. It covers the letters to check, formal and informal address, the difference between Croatian and its close relatives, dates and money, and a short review routine before anything is sent. The examples are written for people who work in Croatian, whether they are native speakers, translators or international teams.

What AI in Croatian does well, and where it slips

Croatian is a South Slavic language with seven grammatical cases, three genders, rich verb aspects and a word order that moves according to emphasis. Assistants handle most of this well because they have seen a great deal of Croatian text. Replies, notices, summaries, product descriptions and polite emails usually come out readable on the first try.

The weak points are predictable. Rare words, legal and technical terms, local institutions and recent changes in rules are less reliable. Case endings after numbers and in long sentences sometimes slip. When your prompt is in English, the assistant may translate word for word and produce sentences that are grammatical but unnatural. And because Croatian shares much of its structure with Serbian and Bosnian, the output can drift toward vocabulary that sounds foreign to a reader in Croatia.

Task	Typical quality	What to check
Short customer replies	Good	Tone, address form, greeting and closing
Summaries of Croatian text	Good	Names, numbers and dates
Translation from English	Good to uneven	Word-for-word phrasing, terminology
Legal and official wording	Uneven	Terms, references and formal structure
Marketing slogans	Uneven	Natural rhythm, wordplay and local feel
Local names and institutions	Uneven	Spelling, case endings and current names

Prompt for natural Croatian

The most useful change you can make is to describe the reader and the situation, and to say which Croatian you want. A prompt such as “Write a reply” leaves too much open. Compare it with: “Write a reply in standard Croatian as used in Croatia to a customer whose parcel arrived late. Use formal address, a polite and calm tone, three short paragraphs, an apology, the new delivery date and a contact line.”

A good prompt contains five parts.

Variety: standard Croatian as used in Croatia, so that words from neighbouring varieties are avoided.
Reader: a customer, a colleague, a municipality, a teacher, with the relationship stated.
Address form: formal (Vi) or informal (ti), chosen by you.
Goal and facts: what must be said, including names, dates, amounts and conditions.
Format: length, headings, bullet points and the closing phrase you want.

Writing the prompt in Croatian often gives more natural output than writing it in English, because the assistant follows the register of your instructions. If you are more comfortable in English, you can still do it: give the instructions in English and paste any Croatian source text as it is. The guide on prompts in languages other than English compares both approaches, and the article on how to write an AI prompt covers the basics.

Diacritics and letters to verify

Croatian uses a Latin alphabet of thirty letters, usually traced to the work of Ljudevit Gaj, and you can read the history of Gaj’s Latin alphabet. Five letters carry marks: č, ć, đ, š and ž. There are also three digraphs treated as single letters: dž, lj and nj. Losing the marks changes meaning or makes a text look careless, so always scan for them.

Most of the time the assistant keeps the marks. Problems appear when text is pasted from a system that strips them, when a prompt itself is written without them, or when a name is copied from an English source. A sentence typed as “cestitam vam” may come back correctly, but a surname such as “Đurđević” can lose its first letter in a form field or a subject line.

Search the draft for words that should contain č, ć, š, ž or đ, and look for suspicious plain c, s, z or dj.
Check that ć and č are not swapped, because both appear in common words and names.
Check that dž, lj and nj were not split into two separate letters in a hyphenated line.
Test the final file, not only the chat window, because email systems, spreadsheets and printers sometimes change encoding.

If you must produce text for a system that cannot show the marks, ask for a second version without them and keep the correct version for your records.

Formal or informal: Vi and ti

Croatian, like many European languages, separates a formal “you” from an informal one. This is called a T-V distinction. In Croatian the formal form is Vi, with the plural verb, and it is capitalised in letters and emails to a single person. The informal form is ti. Using the wrong one in a customer message is one of the most common problems with machine-written text.

As a rule, use Vi with customers, public bodies, older people and anyone you do not know. Use ti with friends, close colleagues and in settings where the culture is informal, such as a young company that addresses its customers casually. When in doubt, choose Vi. It is easier to move to a friendlier tone later than to repair a first impression that was too familiar.

Situation	Suggested form	Sample opening
Customer email	Vi	Poštovana gospođo Horvat,
Offer to a company	Vi	Poštovani,
Message to a colleague you know	ti	Bok Marko,
Social post for a young brand	ti	Bok svima,
Official letter	Vi	Poštovani gospodine Kovač,

Also check the salutation itself. Croatian uses the vocative case when addressing a person, so “gospodin” becomes “gospodine” and “gospođa” becomes “gospođo”. Surnames can change as well. Assistants often get the common forms right and fumble unusual names, so verify them. State the address form in the prompt and ask for it to be kept throughout the text.

Keeping Croatian apart from its neighbours

Croatian, Serbian, Bosnian and Montenegrin are close, and speakers understand one another. Each, however, has its own standard vocabulary and spelling. A text written for readers in Croatia that uses words typical of another variety looks odd, much as an American text sprinkled with British slang looks odd to some readers. This matters in marketing, public information and any message to a customer.

Everyday words: Croatian uses kruh, mlijeko and tisuća, where other varieties use hleb, mleko and hiljada.
Spelling of the reflex of the old vowel: standard Croatian uses the ijekavian forms, such as vrijeme and lijep, whereas ekavian forms such as vreme and lep belong to other standards.
Time and place words: Croatian uses točno where a neighbouring standard uses tačno, and kazalište where another uses pozorište.
International words: Croatian prefers some native terms, such as zrakoplov or glazba, in formal writing, although spoken use varies.

Say “standard Croatian as used in Croatia” in the prompt, and ask the assistant to avoid words typical of other varieties. Then read the result as a local would. If your audience includes the whole region, state that too, because a neutral or bilingual text may be a better choice than forcing one variety. The article on AI for Baltic languages discusses a similar situation, where smaller languages need more careful checking than English.

Numbers, dates, money and names

Formats cause small errors that look unprofessional. Croatia uses the euro, which replaced the kuna at the start of 2023, so old texts and old prompts can contain outdated currency references. Check every amount and ask the assistant to use the euro unless you state otherwise.

Dates: day, month and year with a point after the day, such as 5. studenoga 2026. Months are written in lowercase, and the genitive form is used after the day.
Numbers: a comma separates decimals and a point or a space separates thousands, so 1.250,50 means one thousand two hundred and fifty and fifty cents.
Money: the symbol comes after the number, as in 49,90 €, usually with a space.
Time: twenty-four-hour notation is normal in writing, for example 14:30.
Names and places: check the spelling, the diacritics and the case ending after prepositions.

A few phrases need care when you convert from English: “next Tuesday”, “the end of the month” and “within five business days” depend on context, and an assistant may pick the wrong construction. Give the exact date in the prompt and ask for it to be repeated in the text. For legal deadlines, use official sources, not a machine’s paraphrase.

A review routine for Croatian business replies

A short routine catches most problems before a customer sees them. It takes about two minutes for a typical reply.

Read the first and last lines. Check the salutation, the address form and the closing, such as “S poštovanjem” for formal and “Lijep pozdrav” for friendly.
Scan for diacritics. Look for č, ć, š, ž and đ in names and common words.
Check the facts. Compare every date, amount, name and promise with your source.
Listen for the variety. Replace words that sound foreign to a reader in Croatia.
Read it aloud. Awkward word order and odd phrases are easier to hear than to see.
Send once. Do not paste confidential details into a prompt that you would not send by email.

For anything legal, medical or financial, have a qualified Croatian speaker check the text. Our guide to AI translation quality describes methods for testing a translation, including a back-translation check that can reveal meaning changes.

Using Mio for Croatian work

Mio is one assistant with Chat, Write, Research, Code and Design modes, and you never pick a model: Mio routes each request to one suited to the task. For Croatian, describe the job and the reader in plain words and Mio drafts the text. Write mode suits replies, notices and product descriptions, and Chat mode suits quick translations or explanations of an ending or a word. Research mode, on paid plans, can search the web and cite sources when you need to verify a term or a rule.

Projects and memory help to keep consistency. Store your rules once, such as formal address, the Croatian spelling of your product names and words to avoid, and every new chat starts with them. Your chats and files remain yours, they are not used to train models and you can export or delete them at any time. The free plan includes Chat and Write, which is enough to test the method on real messages.

Frequently Asked Questions

Can AI write good Croatian?

Yes, for most everyday and business writing. Assistants produce fluent, grammatical Croatian in replies, notices and summaries. The risks are subtle: a wrong case ending, a missing diacritic, a word from another variety or the wrong level of formality. Use the assistant for the first draft and let a Croatian reader approve anything customers or the public will see.

Should I write my prompt in Croatian or English?

Either works. Croatian instructions tend to give more natural output because the assistant follows the register of the prompt. English instructions are fine if you paste any Croatian source text unchanged and state the variety, reader and address form clearly. Try both on a real message and keep the version that needs fewer corrections.

How do I keep the Croatian formal?

State it in the prompt: formal address with Vi, polite tone, no slang. Ask the assistant to keep the same form throughout and to use the correct vocative in the greeting. Then check the greeting, the verb forms and the closing line, because a mixed-up form is the most visible error in a customer message.

Why does the output sometimes sound Serbian or Bosnian?

The varieties share most of their structure and a lot of vocabulary, and training text often mixes them. Ask for standard Croatian as used in Croatia, give a few marker words such as kruh, mlijeko and tisuća and read the draft for words that sound foreign. Replace them, or ask the assistant to rewrite the sentence.

Can AI check my Croatian spelling and grammar?

It can find many errors and suggest corrections, and it can explain the rule behind a fix. It can also miss mistakes or change correct text, so review every suggestion. For official documents, publications or legal text, a human proofreader remains the safer final step.

The Bottom Line

AI in Croatian is good enough to save real time, as long as you give it the reader, the address form and the variety, and as long as you check diacritics, case endings, formats and facts before sending. Keep a short review routine, store your style rules once and let a native speaker approve anything important. To try this on a real message, start on the free plan and compare options on the Ask Mio pricing page.

AI in Croatian: Diacritics, Formal Address and Business Replies