The app for independent voices
If you ask an AI, “How do I learn Thai?”, it will give you the internet’s greatest hits: “Try Duolingo” or “memorize common phrases.” None of that is wrong, but none of it is fully aligned with how adult brains actually acquire a new language.
The prompt below fixes that. It hard-codes the evidence-based principles I explored before (sources below this articles) and directs your LLM of choice to build a plan that is realistic, personalized, and actually doable.
All you need to do is copy it, replace the [brackets] with your details, paste it into ChatGPT Deep Research, Claude 4.5 Opus, or Gemini 3.5 Deep Research (I tested all three so you don’t have to), and let it build your roadmap.
The prompt:
# Role
You are an expert researcher and practitioner at the intersection of second language acquisition, AI-assisted learning, and edtech evaluation. You combine knowledge from applied linguistics, cognitive science, and learning design to produce evidence-based, practical recommendations.
# Context
You are advising a motivated learner who wants to learn [TARGET LANGUAGE].
Learner Profile:
Current Level: [Describe your level, e.g., “Absolute zero” or “I did Duolingo for a month but can’t speak”].
Language Background: [List your native language and any other languages you speak proficiently].
Motivation & Timeline: [Describe why you are learning and your deadline, e.g., “I am moving to Tokyo in 6 months” or “I want to speak with my partner’s family by March 2026”].
Specific Interest: [List 2-3 themes you generally like to talk about (this will inform some of the role-playing scenarios) e.g. cultural history, sci-fi books, cooking].
The Definition of Success:
The learner wants to achieve [Target Proficiency Level, e.g., Conversational Fluency / B1 Level] within [Timeframe, e.g., 3 months]. Specifically, they want to be able to:
[Specific Scenario 1: e.g., Order vegetarian food and ask about ingredients]
[Specific Scenario 2: e.g., Discuss my job in Tech Sales with peers]
[Specific Scenario 3: e.g., Navigate public transport and taxis]
# Timeline & Logistics
The plan must fit within these realistic time constraints:
Commute/Passive Time: [e.g., every weekday 30 mins per day walking/driving - audio only].
Daily Review: [e.g., every weekday - 10 mins morning coffee break - phone only].
Deep Work: [e.g., 1x 60-minute sessions every Sunday morning at a desk].
Total Duration: [e.g., 16 weeks].
The learner is willing to use: [e.g., Anki, Spotify, Netflix, “Open to buying 1-2 programs/apps”].
# Constraints & Preferences
The learner values efficiency and evidence-based strategies. Avoid “busywork” (e.g., generic streak apps without substance). Efficiency and evidence-based means:
Do not use traditional classroom methods (rote memorization, isolated grammar drills, or gamified busywork). Instead, structure the entire learning roadmap around the 3-Pillar Framework:
Pillar 1 (Flood the Soil): Focus on Comprehensible Input (95-98% understanding) to lower the affective filter.
Pillar 2 (Grow Grammar): Use “Grammar by Induction” (noticing patterns in context) and occasional explanations rather than memorizing rules.
Pillar 3 (Compost Errors): Use “Errorful Learning” (Output → Feedback → Repair) to turn mistakes into long-term retention.
If the target language uses a different writing system (e.g., Japanese, Thai, Arabic, Russian) and the user is a beginner, consider whether the script is relevant for the desired learning outcome, and if yes, include a specific strategy for learning the script alongside the pillars.
# Task
Design a Language Learning Roadmap. Consider the strengths of Generative AI and Existing Evidence-based Tools (Apps/Courses) to assign the best tool for the job.
For each Pillar, specify:
-> The Strategy: What is the cognitive goal?
-> The “Hybrid” Tool Stack:
Where to use Existing Tools: (e.g., “Use Pimsleur for audio foundations” or “Use Lingvist for high-frequency vocab”).
-> Where to use GenAI: (e.g., “Use Gemini to generate stories about the user's specific interest” or “Use Claude to roleplay a specific scenario close to the desired learning outcome”).
The Routine: A concrete daily protocol mapped against the specified time slots. Ensure the activity fits the device (e.g., no typing exercises during “Phone only” time).
# Output Format
1. The Hybrid Tech Stack
A bulleted list of the 3-4 essential tools selected for this learner, explaining why each was chosen over the others (e.g., “Why Anki instead of a notebook?”).
2. The 3-Month Roadmap (Weekly View)
Break down the [Timeline] into phases.
Phase 1 (Foundation): Focus on...
Phase 2 (Growth): Focus on...
Phase 3 (Fluency): Focus on...
3. The AI Prompt Library
Provide 3 exact prompts the learner needs to copy/paste to make the AI parts of the plan work.
Prompt A (Input Generator): For generating reading material at the right level.
Prompt B (Grammar Detective): For understanding rules via context.
Prompt C (The Simulator): For roleplaying the specific scenarios listed in the Context.
# Meta-Instructions
Balance: Do not force AI where it isn’t needed. If for example, a podcast is better for the commute than an AI conversation with gemini, suggest the podcast.
Efficiency: The plan must fit the Available Time.
Actionable: End with “Day One Instructions”—the exact first thing the learner should do after reading this response.