Two learners talking with no plan usually produces forty minutes of comfortable, easy sentences that neither person learns anything from. A little structure changes the outcome completely.
Finding a speaking partner is the easy part now. Getting value out of the session is where most people fail — two learners meet, talk pleasantly for forty minutes about where they are from and what they do, and leave without having practised anything they could not already say.
Comfortable conversation is not the same as practice. A little structure fixes this, and it costs about two minutes of setup.
Why unstructured sessions underperform
Left alone, two learners naturally drift to the language they share most confidently. That means you spend the session rehearsing what is already automatic. You finish feeling good — fluent, even — and nothing has moved.
Productive practice sits slightly above your comfort level: topics you cannot quite handle, structures you avoid, words you do not yet own. That does not happen by accident.
A session format that works
Forty-five minutes, four parts:
- Warm-up (5 min). Anything easy. Your first minutes are always your worst, and this gets them out of the way before the real work.
- Topic A, your partner's choice (15 min). They lead; you respond and ask questions.
- Topic B, your choice (15 min). You lead. Pick something you cannot discuss easily — that is the point.
- Feedback and repair (10 min). Go through the notes you each took.
The final part is the one everyone cuts, and it is the one that converts talking into learning.
Agree how you want to be corrected
Decide this before you start, because getting it wrong ruins the session in either direction.
- No interruption, notes only. Best for fluency. Your partner writes down errors and gives them to you at the end.
- Correct only what blocks understanding. A good default — the conversation keeps moving but genuine breakdowns get fixed.
- Correct everything. Only for short accuracy drills. Over a long conversation it destroys flow and confidence.
Tell your partner which one you want, and ask what they want. Mismatched expectations — one person correcting every article while the other wanted to practise flow — is the most common reason a partnership fails.
Set one narrow target per session
"Improve my English" is not a target. These are:
- Use the present perfect at least ten times.
- Use five specific phrasal verbs from this list.
- Never use the word "very" — find a stronger adjective each time.
- Ask at least eight questions.
- Do not switch to my own language once.
A narrow target gives your partner something concrete to watch for, and gives you a clear pass or fail that does not depend on how you felt.
Topics that force real language
Biographical questions produce rehearsed answers. These do not:
- Explain how something works — your job, a process, a recipe. Forces sequencing and precision.
- Compare two things and pick one — forces comparatives and justification.
- Tell a story that went wrong — forces past tenses under load.
- Argue a position you disagree with — forces opinion language and hedging.
- Describe a photo neither of you has seen — forces articles and vocabulary (surprisingly effective, see a, an and the).
- Plan something together — a trip, an event. Forces future forms and negotiation.
Language exchange: protect your half
If your partner is learning your language, agree on a hard split — 20 minutes each, timer on. Without it, the session slides into whichever language is easier, and the less confident speaker loses. Switching languages every few sentences feels friendly and teaches neither person very much.
Taking useful notes
Do not write everything. During the session, capture only:
- Moments you got stuck — what you wanted to say and could not.
- Words you used in your own language because you lacked the English.
- Anything your partner said that you want to steal.
That first category is the most valuable material you will ever generate, because it is a syllabus written by your own gaps. Look those up afterwards and deliberately use them next time — the production step described in building vocabulary that sticks.
Choosing a partner
Level matters less than reliability and attitude. A partner slightly above you is ideal, but a committed partner at your exact level beats an advanced one who cancels. Look for someone who shows up, who is willing to correct you, and who can take correction without defensiveness.
One warning: a partner who only wants to chat socially is not a practice partner. That is fine — friendship is valuable too — but be honest with yourself about which one you have, and find a second person if you need actual practice.
When a human partner is not available
The obvious limitation of partners is scheduling. An AI tutor has no schedule, never gets bored of your mistakes, and will happily run the same drill ten times — which makes it better than a human for repetition and targeted correction, and worse for the unpredictability that real conversation provides.
Use both. Drill with AI, then take what you drilled into a room with real people. On Zeno you can do both in one place: Voice Mode for private repetition, live Spaces for the messy real thing. Before joining your first Space, the Community Guidelines explain how to report or block anyone who makes the room uncomfortable.
Finally, agree the next session before you leave this one. Partnerships die in the gap between "we should do this again" and actually scheduling it.