Writing Evidence Card
Good writing is not only attractive sentences. It completes a task that can be explained, checked, and answered. Copy this card into a private note and keep the draft, revision record, and delivered version. Do not record passwords, client data, unauthorised third-party information, or unnecessary personal details.
1. Define the task and boundary
# Writing Evidence — YYYY-MM-DD
Real context:
Task: explain / decide / narrate / reflect
Audience and existing context:
Delivery channel and deadline:
Action the reader must be able to take:
Content that must stay private:| Acceptance condition | How to check |
|---|---|
| Reader understands the point | Ask one reader to restate it in one sentence |
| Information can be checked | Mark numbers, dates, quotations, and sources |
| Structure is usable | Reader can find conclusion, reasons, and next step |
| Boundaries are respected | Check privacy, copyright, conflicts, and necessary disclosure |
2. Save an unaided draft
Draft location:
Start and finish time:
Sentence I most want to keep:
Place I am least certain about:
Draft completion standard:Write the first version yourself without AI writing it. It is the comparison sample for future growth.
3. Log four revision passes
| Version | Revision layer | What I changed only | Why | Evidence/feedback |
|---|---|---|---|---|
| v1 | Purpose and facts | |||
| v2 | Structure and evidence | |||
| v3 | Sentences and voice | |||
| v4 | Delivery and boundaries |
Keep the order: purpose and facts → structure and evidence → sentences and voice → delivery and boundaries. Address the highest-impact problem in each pass instead of treating edit count as quality.
4. Record AI and tool involvement
Tool/model and date:
Task I gave it:
Three problems it raised:
What I accepted/rejected, and why:
How I checked facts, citations, privacy, and copyright:
What I can still explain independently after closing the tool:AI can diagnose structure, simulate a reader, or generate parallel exercises. The author remains responsible for facts, tone, permission, and final judgment.
5. Ask for specific feedback
Ask readers questions tied to the task:
- What did you think the piece wanted you to understand, decide, or do?
- Where did you pause or have to guess?
- Which fact, quotation, or link needs checking?
- Which sentence would you delete, and why?
| Reader | Restated gist | Largest barrier | Change I will test |
|---|---|---|---|
6. Test delayed transfer and delivery
| Date | Changed condition | Output | Result |
|---|---|---|---|
| Day 1 | Original task, timed, unaided draft | Draft and four revision passes | |
| Days 3–7 | Related topic or new audience | 200–300 word parallel text | |
| Day 30 | New topic or real delivery | Email, report, explanation, or long-form piece |
Score 0–2 for purpose, structure, evidence/detail, accuracy/range, and revision transfer. Strong performance only on a familiar topic or after tool polishing is not full transfer.
7. Decide the next step
The strongest part of this piece:
The largest remaining task barrier:
One variable I will practise next:
Delivery link or archive location:
Next review date:Related chapter: Writing: From Draft to Verifiable Revision