A normal chat with an AI hands you the answer and stops there — for learning a language that isn't enough, you need someone who asks you questions back and checks that you understood. Gemini has a mode for exactly this: Guided Learning turns the model from a reference book into a conversation partner that follows a plan and adapts the difficulty to your answers. Below is how to switch the mode on and which prompts give a predictable result at each step.
The idea: dialogue instead of a lecture
The difference between "explain the rule" and Guided Learning is that in the second case the model asks clarifying questions before giving you material, and then checks that it stuck — through questions and error analysis, not through simply laying out the theory. It's the same principle good learning prompts are built on in general: not "tell me", but "ask me, then teach me, then test me".
Step 1. Switch on Guided Learning
Open gemini.google.com (or the Gemini app), find "Guided Learning" in the chat tools under the input box and select it. After that the model stops answering questions straight away and starts clarifying the context — that's the expected behavior of the mode, not a glitch.
Step 2. Set your level and goal in a single prompt
For Gemini to build a personal plan rather than a generic course, your first message should state your starting level, target level, native language and timeframe right away:
I want to learn English from A2 to B1. My native language is Russian,
I have 30 days. Build me a personal day-by-day study plan,
test me at every step and adapt the difficulty
to my answers.
What matters is putting in your own specifics: your real current level (if you're not sure, ask Gemini to assess it first with a short check), a real timeframe and what you want to focus on (grammar, speaking, vocabulary for work) — without that the plan comes out too generic.
Step 3. Work through grammar in dialogue, not by rote
Instead of asking "explain the Present Perfect", ask the model to check your understanding right away rather than just handing you the rule:
Explain the Present Perfect in plain words. Give 3 examples
from everyday life, then ask me 5 questions to check.
If I get one wrong — explain the mistake and give me
a similar exercise.
This prompt works for any grammar topic — only the name of the tense or structure changes. The model won't formulate the correct answer for you: it makes you produce it yourself, and that's closer to how memory actually works when learning a language.
Step 4. Practice speaking through Gemini Live
For spoken practice, switch on the Gemini Live voice mode and set a role-play situation:
Let's role-play. You're a barista in a cafe in London, I'm a tourist.
Have the conversation in English, correct my mistakes gently
and keep the conversation going.
The role can be changed to fit the task: a job interview, a conversation with a doctor, small talk with a colleague — the principle is the same. The model stays in character and corrects mistakes gently as it goes, without breaking off the conversation entirely.
Step 5. Upload material to work through
Gemini can work with uploaded files inside the same conversation:
- A textbook PDF → a short summary and a quiz on the material.
- A screenshot of a chat in English → a breakdown of the structures and a translation.
- A YouTube video → a transcript with the difficult phrases explained.
- A photo of a menu or a sign → a translation and an explanation of the grammar in the text.
It's handy to do this as a continuation of the same conversation where your level and goal are already set — then the material is broken down at your current difficulty rather than generically.
The full sequence
- Switch on Guided Learning in the chat tools.
- State your level, timeframe and goal in a single message.
- Work through grammar topics with the "explain → examples → test me → break down my mistake" prompt.
- Supplement text practice with the Gemini Live voice mode and role-play dialogues.
- Periodically upload your own material (PDFs, screenshots, videos) to be worked through at your level.
How we do it
The same principle — role, task, step-by-step checking — underpins the learning prompts and skills collected in the qvib.pro knowledge base: how to build role-play dialogues in Claude, how to break down uploaded material with agentic skills, and how to construct a teaching scenario for any subject, not just languages. No invented reach figures and no promises — only verifiable prompts and open access, with no code words and no DMs.
Related links
- Prompts for learning scenarios: qvib.pro/arsenal/prompts
- Claude skills for working through material: qvib.pro/arsenal/claude-skills
- Vibe coding tools: qvib.pro/docs/vibe-coding-tools