Which tool to choose

All the main AIs speak Italian, but some slip into English more easily, because they were trained mostly on English texts.

  • If the AI replies sometimes in English on certain technical topics, it's inertia: the topic has more material in English and the AI falls back into it. It's corrected with a firm instruction.
  • If the AI replies almost always in English despite you writing in Italian, try setting the language in its options, if it has them, or choose an assistant that handles Italian better. For someone who works in Italian all day, it's worth looking for one that doesn't need correcting constantly.

Before changing tools, though, an explicit request to reply in Italian solves the vast majority of cases. Below you'll find how to make it effective.

How to do it

From a computer or a phone the method is identical: make the language an instruction, not an expectation.

  1. Ask it clearly. Don't take for granted that it follows you because you write in Italian: say it. The operating syntax:
Always reply to me in Italian, even when I paste you texts in other
languages or we talk about technical topics. If you have to use an
English term with no translation, keep it in English but write everything
else in Italian.
  1. If you happen to paste it a long text in English, reiterate the language in the same message: "below there's a text in English, but you reply to me in Italian". The English text is the most common cause of the relapse.
  2. If it falls back anyway, correct it right away, in the next message: "you went back to English, continue in Italian". The sooner you catch it, the sooner it settles.
  3. For long jobs, check in the AI's settings whether there's a language preference or a "custom instructions" section where you can set Italian once and for all.
  4. Check: after the instruction, the next two or three replies should be in stable Italian. If they are, it has locked onto the language; if it falls back again, there's an active cause (a term, a text) to neutralize.

A concrete example

Elena asks the AI to explain a computing concept to her. The reply comes entirely in English, because on that topic the training material is predominantly English and the AI slips into it.

Elena replies: "in Italian, please, and from now on always in Italian even on technical topics". The AI redoes the explanation in Italian and continues that way for the whole conversation. It wasn't a malfunction: it was the inertia of the topic's dominant language, and a clear instruction was enough to overcome it.

When it does NOT work (and how to fix it)

If it falls back into English every time you paste a foreign text

It's the number one cause. When you feed it a block of English text, the AI tends to "think" in English and reply to you that way. The fix is to anticipate it in the same message: "the text below is in English, analyze it but reply to me in Italian". You say it before, not after receiving the wrong reply.

If it mixes Italian and English in the same sentence

It happens with technical terms. If it bothers you, ask it to translate those too: "use the Italian terms when they exist, and when you have to keep an English word put in parentheses what it means". If instead the technical English terms are fine with you (sometimes they're clearer), leave them and correct only the rest.

If it goes back to English after a while of going in Italian

The effect of the instruction fades in very long conversations. A reminder is enough: "remember, in Italian". For long jobs, repeat the language instruction now and then, or put it in the permanent settings if the AI offers them, so you don't have to remind it anymore.

If even with the explicit instruction it continues in English

Then that specific AI handles Italian badly, or there's a system language setting set to English. Check the account options and the interface language. If it's all in order and it insists, consider a different assistant: for someone who works in Italian, fighting every reply is a waste of time that another tool saves you.

A tip from someone who actually uses it

If you use the AI often in Italian, look once for the permanent instructions section (many assistants have one, called "custom instructions" or similar) and write in it "always reply in Italian". Setting it there applies to all future conversations and frees you from having to repeat it every time. Five minutes spent once remove an annoyance that otherwise comes back with every chat.

Frequently asked questions

Why does the AI "prefer" English?

Because most of the texts it was trained on are in English. It's its starting language, the one in which it finds the most material, especially on technical topics. It replies in Italian perfectly well, but if nothing anchors it to our language it tends to go back toward the one it "read" the most in. The explicit instruction is the anchor.

Do I have to write in English myself to get better answers?

Not for the quality of the Italian language, which is excellent. On very technical and niche topics, sometimes a question in English fishes out slightly richer information, because the material is abundant there. But for normal use go ahead and write in Italian: the difference, if any, is minimal and not worth the inconvenience.

Can I have it translate into Italian a reply it already gave in English?

Yes, you just have to ask: "translate into Italian the reply you just gave me". It does it instantly and keeps the content. In fact, it's a useful trick: you let it reason as it prefers and then have the result delivered in your language.

Does replying in English mean it understands Italian less?

No, and it's the misunderstanding to clear up. The AI perfectly understands what you write in Italian: the proof is that it answers on point, just in the wrong language. Replying in English isn't a comprehension problem, it's a linguistic habit problem, like a bilingual person who, distracted, replies in the language they were thinking in. You understood perfectly what it asked you; it just got the language wrong in replying. And like that person, telling it is enough to make it change.