Skip to content

Teaching Personas

A Teaching Persona is a named set of instructions that shapes how the AI generates lesson content. Different learners may benefit from different tones, structures, or pedagogical approaches — personas let you define those once and reuse them.

Creating a Persona

Go to Teaching Personas in the sidebar and click Add Persona. Fields:

Field Purpose
Name A short label (e.g. "Interest-Led Coach", "Structured Direct")
Instructions Free-text prompt instructions added to every AI generation using this persona

Assigning Personas to Learners

Each learner can have a Default Persona. This persona is pre-selected whenever the AI Generator is configured for that learner. You can override the persona on a per-generation basis in the AI Generator config.

The Instructions field accepts any free text, but the personas that produce the most consistent results follow a three-part structure:

  • [Role] — who the AI is and what it specialises in. Anchors the AI's voice and stance for every generation.
  • [Core Directives] — the pedagogical rules the AI must follow when generating content. Mix philosophy ("frame instruction as a solution to a problem the child is facing") with operational rules ("cap direct instruction at 15 minutes"). This is where follow-up cadence and spaced-repetition guidance belong (see callout below).
  • [Output Requirements] — content that every lesson must contain, regardless of which output template is used. Use this for things like "must include a multisensory element" or "must include a real-world utility link".

This structure works because it separates who the AI is (Role) from how it should think (Core Directives) from what it must produce (Output Requirements). Each section answers a different question, so the AI is less likely to drop one set of constraints when honouring another.

!!! note "Follow-up cadence belongs in [Core Directives]" The prompt automatically tells the AI which prior lessons exist, the date of the earliest achievement in the visible window for each one (e.g. "earliest recent achievement 5 days ago"), and the full list of achievements that followed. A directive like "Schedule review sessions on an expanding interval — 24 hours, 3 days, 1 week, 1 month after first exposure" lets the AI reason about which prior lessons are due for a follow-up. Without a directive, the AI has the data but no rule for when to act on it. (The label is "earliest recent achievement" rather than "first exposure" because the true first-exposure date may lie outside the achievement lookback window.)

Worked Example: Interest-Led Coach

[Role]
You are an expert teaching coach assisting a homeschooling parent. You specialize in novel, interest-led education that integrates a child's passions to keep learning engaging and effective.

[Core Directives]
When generating lesson ideas, schedules, or advice, you must strictly adhere to the following pedagogical rules:

1. Provide "Just-in-Time" Instruction: Design lessons using the "Minimal Effective Dose." Cap direct instruction at 15 minutes. Frame instruction as a solution to a problem the child is currently facing in their play or interests.
2. Emphasize "Strewing" (Self-Directed Learning): Honor the child's autonomy by suggesting materials, games, or books to "strew" in their environment rather than forcing seated instruction. Allow academics to progress in leaps.
3. Utilize "Incidental" Therapy: Do not suggest isolated drills (like "eye push-ups"). Always embed therapeutic or foundational skills (like visual tracking) into goal-oriented, functional play (like playing Zingo or Rummy).
4. Design for Spaced Repetition: Schedule review sessions using an expanding interval schedule (e.g., 24 hours, 3 days, 1 week, 1 month). Ensure a full sleep cycle occurs between the first and second exposure to new material.
5. Enforce Desirable Difficulty: Suggest active recall methods rather than passive reading.
6. Cultivate a Growth Mindset: Always frame feedback and learning goals around effort and growth rather than performance or competition.

[Output Requirements]
Every lesson plan or schedule you generate must include:

Multisensory Elements: Explicitly state how to use tools (e.g., Sand Tray, manipulatives).

Utility-Links: Provide a clear, real-world reason for the child to learn the skill.

Educational Benefits Summary: Clearly summarize the academic, social, and extra-mural benefits of the suggested activities.

Household Tasks: Suggest age-appropriate chores that integrate with the learning goals and make a meaningful contribution to the family or wider community.

Why this works:

  • The [Role] line is short and concrete. It names the expertise ("teaching coach"), the audience ("a homeschooling parent"), and the stance ("interest-led, integrate passions"). Every lesson the AI generates will inherit that framing without you having to repeat it.
  • Directive #4 ("Design for Spaced Repetition") is the directive that makes the achievement section of the prompt useful. The prompt surfaces the "earliest recent achievement N days ago" timestamp on every prior lesson — directive #4 tells the AI what to do with that information. If you remove this directive, the AI still sees the dates but has no rule for when to schedule a follow-up.
  • The [Output Requirements] block layers on top of whatever output template you select. The template controls the shape of the lesson (e.g. "30-minute structure with warm-up, main, cool-down"); the Output Requirements ensure mandatory content (multisensory, utility-link, household-task) appears regardless of which template is in play.

Writing Your Own

Use this checklist when drafting a new persona:

  • [Role]: 1–3 sentences. Name the AI's expertise (coach, tutor, Socratic guide…), the audience, and the stance. Avoid abstract adjectives ("excellent", "knowledgeable") — name a style instead.
  • [Core Directives]: Numbered list. Each directive is one rule. Mix pedagogy with operations. If you want follow-up lessons, add a spaced-repetition directive with explicit intervals (e.g. "24h, 3 days, 1 week, 1 month") — the AI can see lesson dates and "N days ago" labels in the prompt, but only acts on them if a directive tells it to.
  • [Output Requirements]: Bullet list of content that must appear in every lesson, regardless of the chosen output template.

How Personas Combine with the Prompt

When a persona is selected, its name and instructions are injected directly into the prompt sent to the AI under a ### TEACHING PERSONA heading, near the top — above the learner profile, prior lessons, and achievement history. The persona is the AI's standing instruction set for everything that follows.

Personas pair with output templates: the persona shapes the AI's voice and pedagogy; the template shapes the structure and length of the generated lesson. They are independent and complementary — see Output Templates for the template side.