An AI Spanish tutor: what it notes and what it cannot do

Updated 2026-10-18

In Convola, the AI tutor is separate from the people you talk to. It reads each of your lines, notes up to three mistakes in English, and sends them into practice. Here is what it noted in one recorded conversation, how a note comes back, and where it falls short.

Two kinds of AI in one scene

The characters in a scene, a waiter or a landlord, are AI. They answer in Spanish and stay in the scene. They do not stop to correct you.

A separate AI tutor reads your lines while the scene goes on. Its notes appear on your message as small marks you can open. The conversation keeps its flow, and every correction is still there when it ends.

Both use the Spanish of Spain. Speaking to a group as vosotros is right there, not a mistake.

What it noted in a recorded conversation

From a recorded Convola conversation: the A1 scene of a hotel breakfast with a chatty guest, set in Seville. The learner wrote two lines. The tutor marked three things.

Each note names a topic (here ser vs estar, questions and noun gender), the wrong part, the right one and a reason in English. Grammar topics come from a fixed list, so the same mistake is filed in the same place each time.

How a note comes back

Each note goes on your list of issues and returns in practice. For the first note above:

1. In the conversation: Soy de vacaciones is marked, with Estoy de vacaciones as the fix.
2. At the next practice: a gap to fill, ___ de vacaciones en Sevilla, with the verb given as a clue.
3. Days later, a new prompt: say in Spanish, we are on holiday with friends. Estamos de vacaciones con amigos.
4. Right twice, days apart, and the issue leaves your list. A new slip brings it back.

A mistake whose fix is a fact, such as a noun's gender or a verb form, also becomes a grammar card. Open issues are also woven into later scenes, so you meet them again in conversation. You can ask for a short lesson in English on a grammar topic it noted.

At higher levels

From C1 the tutor also flags Spanish that is correct but not what a native would say. It files these under register, collocation, idiom, precision, naturalness and discourse markers. Real mistakes still come first.

Below C1 these style notes are left out, so a beginner is not buried in fine points. The scenes get harder too: negotiating a pay rise with a manager on a tight budget, or a radio debate where the host keeps cutting you off.

What it cannot do

• Be right every time. The tutor is an AI model, and so are the characters. A note can miss a mistake or mark something that was fine. If one looks odd, check it in a grammar reference.
• Catch everything in a long line. It notes at most three things per line.
• Pick on choices a native would also make. What an educated native speaker would say is not marked, since spoken Spanish often has more than one right answer.
• Hear your accent. Your speech is turned into text, and the tutor reads the text, so it does not judge pronunciation.
• Know you as a teacher does. It sees your open issues, not your goals or your week. A human tutor can try a new explanation when one does not land.

When a person is the better choice

A human tutor suits you if you want someone to plan a course, correct your pronunciation by ear, or talk at length about your own topics. A conversation exchange gives you a real person and their culture.

An AI tutor suits practice at odd hours, and the moments you would rather nobody watched you get it wrong. They do different jobs, and you can use both. For the slips English speakers often make, see common Spanish mistakes.

More Spanish guides

• Common Spanish mistakes for English speakers, and how to fix them
• How to practise speaking Spanish by yourself
• All Spanish guides