All articles

Your AI Language Coach

A modern AI like Claude, ChatGPT, or Gemini is the closest thing to a free, 24/7, judgment-free personal trainer for your target language — if you tell it what to do. This page hands you the prompt pack so it generates today's exact session, finds you i+1 input, and spots you through conversation reps.

5 min readLanguide Wiki
On this page
A modern AI like Claude, ChatGPT, or Gemini is the closest thing to a free, 24/7, judgment-free personal trainer for your target language — if you tell it what to do. This page hands you the prompt pack so it generates today's exact session, finds you i+1 input, and spots you through conversation reps.

What it is

Think of an AI chatbot as a personal trainer who never sleeps, never sighs at your bad accent, and costs roughly nothing. A coach's whole job is to look at where you are, hand you the right work for today, and keep you coming back. That's exactly what a well-prompted LLM does.

Most people use AI for languages the lazy way — "translate this," "give me a grammar rule" — and get the same disappointing gym-bro advice everyone gives: memorize 5,000 words, cram verb tables, "speak from day one." That's the equivalent of doing bicep curls in the mirror and wondering why you can't run a marathon.

Used correctly, AI does something a textbook never could: it generates infinite material at your exact level on whatever you actually care about. You don't adjust to the curriculum — the curriculum adjusts to you. Bored of restaurants and train stations? Get a beginner-level dialogue about skateboarding, your job, or zombie apocalypses instead. That's not a gimmick. Interesting, understandable input is the entire engine of acquisition, and AI is an input-generating machine.

The deal here is simple: you bring the consistency, the AI brings the programming. This page gives you copy-paste prompts that turn a generic chatbot into a coach that knows your level, your minutes-per-day, your interests, and what to push you on next.

The evidence

The science behind this is older than the chatbots. Linguist Stephen Krashen's Input Hypothesis argues we acquire language by understanding messages slightly beyond our current level — what he calls i+1 (your current competence, i, plus a small, comprehensible step up). Not by memorizing rules. Not by drilling output before we're ready. By understanding stuff.

Krashen also described the affective filter: when you're anxious, embarrassed, or stressed, a mental "filter" goes up and blocks acquisition even from good input. Low stress = filter down = the input actually lands. (See Affective filter.)

AI happens to be almost custom-built for both ideas:

  • Infinite i+1 comprehensible input. Ask for a story "for an A2 learner, 80% words I already know" on any topic and you get exactly that — endlessly, on demand. This solves the classic beginner problem: real native content is too hard, and graded textbook content is boring. (See The Input Hypothesis (i+1) and Finding comprehensible input.)
  • A near-zero affective filter. There's no human watching you fail. You can mangle a sentence, ask the same "dumb" question nine times, and request things slowly with no judgment. For shy learners this is the unlock.
  • Instant, on-demand feedback. Confused by one sentence? Ask. The explanation appears in seconds, in your native language, at your level.

Now the honest part — because a good coach doesn't lie to you:

  • AI hallucinates. It can invent a "grammar rule," give a slightly off translation, or confidently produce a phrase no native would say. Treat its grammar explanations as a helpful first draft, not gospel. Cross-check anything important against real native content or a reference.
  • AI is text-first. Even with voice modes, its accent and rhythm are not a reliable model for pronunciation. Get your sound from native speakers — shows, podcasts, real people. AI is your sparring partner, not your pronunciation idol.
  • AI is a tool, not the territory. It can't replace hours spent inside the actual language. It makes those hours easier to start and better aimed. The reps still have to be yours.

Bottom line: AI is a phenomenal delivery mechanism for comprehensible input and a tireless conversation partner — as long as you verify its facts against the real thing.

How to actually use it

This is the prompt pack. Paste these into Claude, ChatGPT, or Gemini. Swap the [PLACEHOLDERS]. Reuse them every day.

The system runs on a small Training Code the AI generates once and you paste into every other prompt — like a profile on a gym app, so the coach never forgets your level.

(a) The Diagnostic — generates your personal Training Code. Run this first, once.

` You are my no-nonsense language coach. I want to learn [LANG]. My native language is [NATIVE LANG]. Interview me with 6 short questions to gauge my real level — ask one at a time, wait for my answer, and adapt. Cover: how much I understand when listening, reading, my goals, my interests, and how many minutes per day I can realistically train. Be honest, not flattering.

When done, output a single line called my TRAINING CODE in exactly this format: LANG=[language] / LEVEL=[A1-C2 estimate] / MIN=[minutes per day] / DIET=[my interests, comma-separated] / FIRE=[the #1 reason I'm doing this]

Then give me one sentence of brutally honest assessment of where I am. `

(b) The Daily Session Generator — reads your Training Code, hands you today's workout.

` Here is my TRAINING CODE: [PASTE TRAINING CODE]

Act as my coach. Build me ONE session that fits in MIN minutes, using INPUT-FIRST principles (comprehensible input is the priority; do NOT make me memorize word lists or grammar tables). Structure it as:

  1. WARM-UP: 3-5 sentences at my LEVEL on a topic from my DIET, with a tiny glossary for new words.
  2. MAIN INPUT: a short story or dialogue (~150 words) at roughly i+1 — 80-90% words I likely know,

on a DIET topic. Then 3 simple comprehension questions in [LANG].

  1. COOL-DOWN: one optional 2-minute output rep (only if I feel ready) tied to my FIRE.

Keep instructions in my native language; keep the [LANG] content at my level. End with one line of encouragement that references my FIRE. No fluff. `

(c) The Comprehension Companion — explains a sentence without a grammar lecture.

` My TRAINING CODE: [PASTE TRAINING CODE] Here is a sentence in [LANG] I don't fully understand: "[PASTE TEXT]"

Explain what it MEANS in plain [NATIVE LANG], like a coach pointing at a screen — not a textbook. Give: (1) a natural translation, (2) a quick gloss of any 1-2 words I probably didn't know, (3) ONE intuitive note on why it's built that way, in a single sentence. Do NOT lecture me on grammar terminology. Do NOT give me a verb table. Keep it under 80 words. `

(d) Sentence-Mining → Anki card maker. Turns input you've understood into review cards.

` My TRAINING CODE: [PASTE TRAINING CODE] Here is text in [LANG] I just read/watched and mostly understood: "[PASTE TEXT]"

Mine 3-5 of the most useful sentences for me to remember (prioritize high-frequency, reusable phrasing over rare words). For each, output a row I can paste into Anki, tab-separated: FRONT = the full [LANG] sentence with ONE target word in {{c1::cloze}} brackets BACK = natural [NATIVE LANG] translation + a 1-line note on the target word Output ONLY the rows, nothing else, so I can paste straight into a CSV/Anki import. `

(More on this technique: Sentence mining and Spaced repetition.)

(e) The Conversation Partner — sparring with gentle correction.

` My TRAINING CODE: [PASTE TRAINING CODE]

Be my conversation partner in [LANG], staying at my LEVEL (simplify if I struggle). Pick a scenario from my DIET and start the conversation — keep your messages short. Rules:

  • After EACH of my replies, give me a quick "spot check": if I made a mistake that blocks meaning,

gently restate the correct version in ONE line, then keep the conversation going. Ignore tiny errors — don't kill the flow or my confidence.

  • If I freeze, offer me two example things I could say.
  • Keep my affective filter LOW: encouraging, zero judgment, never sarcastic about errors.

Begin now with the scenario and your first message. `

Don't force this before you're ready — output emerges naturally once you've banked enough input. If it feels miserable, you need more reps from prompts (b) and (c) first. (See The speaking skill.)

(f) The Plateau-Buster / Weekly Review — when progress stalls.

` My TRAINING CODE: [PASTE TRAINING CODE] This week I did: [describe honestly — e.g., "3 sessions, watched 2 episodes, skipped conversation, felt stuck understanding fast speech"].

Act as my coach doing a weekly review. Give me:

  1. One honest diagnosis of WHY I feel plateaued (be specific — usually it's not enough input, or

input that's too easy/too hard, or avoiding listening).

  1. ONE concrete change to my next week, INPUT-FIRST (more/harder/more interesting input — not grammar

drills, not "fluent in 30 days" nonsense).

  1. An updated TRAINING CODE if my LEVEL or DIET should change.
  2. One sentence connecting this back to my FIRE so I actually show up. No motivational clichés.

`

A real plateau is almost always an input problem in disguise — usually too little, or input that stopped being i+1 and got comfortable. The fix is rarely "more grammar."

Resources

Real tools, no fake links:

  • Claude (claude.ai) — strong at long, nuanced explanations and staying in character as a coach; great for the diagnostic and conversation prompts.
  • ChatGPT (chat.openai.com) — solid all-rounder; its voice mode is handy for low-pressure speaking reps (just don't trust its accent).
  • Gemini (gemini.google.com) — good free tier, integrates with Google search for verifying claims.
  • Language Reactor — browser extension for Netflix/YouTube that shows dual subtitles and lets you mine sentences from real native video. Pair it with prompt (d): mine a line, paste it in, get an Anki card.
  • Anki — the free spaced-repetition app. Import the cards from prompt (d) and review daily. See Anki guide.

The killer combo: native video (Language Reactor) → understood sentence → AI turns it into a card (prompt d) → Anki reviews it. Input, mined, locked in.

Gear on the flywheel

The stuff that actually moves your reps

Real resources for this page — ranked by learners, never sponsored. Tap through to upvote, save, or grab them.

Keep going — Start Here

The rest of this shelf. Pick the next rep.