AI Disclosure and Authentic Work

Using AI on your SAT is allowed — not disclosing it isn't. This page explains what to log, why it matters, and how validation catches students who skipped this step.

flowchart TD
    A[You need to do C01 work] --> B[Use AI — ask it something]
    B --> C[Read the output carefully]
    C --> D{Is it right for<br/>your project?}
    D -- Yes, mostly --> E[Keep what fits,<br/>rewrite the rest]
    D -- No / generic --> F[Reject or heavily<br/>rewrite it]
    E --> G[Open AI-disclosure-log.md<br/>Add an entry: prompt · output · kept · changed]
    F --> G
    G --> H[Commit to GitHub]
    H --> I[Validation — locked conditions<br/>No AI. No phone. No internet.]
    I --> J{Can you explain<br/>your decisions?}
    J -- Yes --> K[You pass the<br/>authenticity test]
    J -- No --> L[Gap between your brief<br/>and your understanding shows]

    classDef action fill:#dbeafe,stroke:#2563eb,color:#1e3a8a
    classDef decision fill:#fef3c7,stroke:#d97706,color:#78350f
    classDef keep fill:#dcfce7,stroke:#16a34a,color:#14532d
    classDef reject fill:#fee2e2,stroke:#dc2626,color:#7f1d1d
    classDef log fill:#ede9fe,stroke:#7c3aed,color:#4c1d95
    classDef validate fill:#fef2f2,stroke:#991b1b,color:#7f1d1d,stroke-width:2px
    classDef pass fill:#bbf7d0,stroke:#15803d,color:#14532d,stroke-width:2px
    classDef fail fill:#fecaca,stroke:#b91c1c,color:#7f1d1d,stroke-width:2px

    class A,B,C action
    class D,J decision
    class E keep
    class F reject
    class G,H log
    class I validate
    class K pass
    class L fail

Why disclose?

Your teacher compares what you wrote with what you say. A strong brief paired with vague answers under questioning signals that the brief doesn't reflect your thinking. A logged AI interaction, by contrast, shows you engaged critically — you read it, decided what to keep, and can explain why.

The assessment is designed so a student who outsourced their thinking cannot fake it at validation. Presentation, writing test, and interview all happen under locked conditions (no AI, no internet, no phone). Your log is also your study notes: it records what you actually decided, so you can recall and defend it.

What to log

Every C01 AI interaction gets an entry in AI-disclosure-log.md in your C01/ GitHub folder:

## Entry — [Date] — [Task it relates to]

**What I asked the AI:**
[Paste or summarise your prompt]

**What it produced:**
[Paste the key output, or summarise if long]

**What I kept:**
[What went into your work unchanged or nearly unchanged]

**What I changed or rejected:**
[What you rewrote, cut, or decided not to use — and why]

Common C01 uses that must be logged: drafting any part of your design brief, generating Gantt chart task lists, understanding PSM stages, generating Mermaid or Excalidraw code, reviewing your "Why monitor" response.

Commit the log each time you add an entry — the git history shows you kept it current, not that you backfilled it the night before.

The AI-first workflow

The flowchart above is the expected workflow. AI is a starting point, not a final answer. What matters at validation is that you can explain every decision in your brief — why you framed the problem that way, why you chose those users, why that language. If AI suggested something and you kept it, you should be able to say why it was right for your specific project, not just that "AI said so."

At validation: where authenticity is tested

C1 uses three separate assessment formats — presentation, writing test, and interview — all conducted under locked conditions. No AI. No phone. No laptop beyond what's permitted. A student whose brief was largely AI-generated cannot answer probing questions like "What makes your solution original?" or "What would you do if your chosen library stopped being maintained?" with any specificity. Your spoken answers are compared against your written work; consistency across both is what earns marks.

Common mistakes

Writing "I used AI to help" with no detail about what you asked or what changed — this is not a valid entry and suggests you have something to hide.
Backfilling the log in one batch before submission — the commit timestamps will show it; add entries as you go.
Logging that you "kept everything" without explanation — if you kept the full AI output unchanged, explain concretely why it was exactly right for your project.
Forgetting that paraphrasing counts — if AI wrote 300 words and you reduced it to 80, that still needs an entry.