Draft a useful article from a factual brief

Keep a source ledger and an explicit “do not invent” list alongside the writing prompt.

Workflow guideEditorial update 2026-10-13

Tools for this workflow

ChatGPT, Claude, Grammarly

Start with this input

Audience: independent tutors.
Topic: preparing a first lesson.
Known steps: ask the learner’s goal; run a short baseline task; agree one measurable target.
Do not invent research statistics or promise exam gains.
Deliver: a 3-section outline and a 70-word introduction.

Follow the process

  1. Copy the supplied brief into a new conversation. Ask for an outline before requesting prose.
  2. Check that every section serves the audience and that no result, statistic or testimonial was added.
  3. Draft one section at a time. Ask the model to mark any claim that needs a source with [VERIFY].
  4. Open primary sources for any new factual statement. Remove claims you cannot substantiate.
  5. Edit for your own voice, then copy into your publishing tool and inspect headings and links before publishing.

Reference output and acceptance criteria

Reference outline:
1. Learn what success means to this learner
2. Use a short baseline task
3. Agree one observable target

Reference opening: A first lesson is a chance to discover what the learner needs next. Ask what they want to do, watch them attempt a small task, then agree one target you can both recognize. A clear starting point makes the next lesson easier to plan.

If it goes wrong

If prose adds an unsupported claim, return to the brief rather than asking for a more confident answer. If the tool hits a quota, save the draft locally and continue editing manually.

Time, cost and evidence

Planning estimate: allow 15–30 minutes for a short first run and manual review. This is not a measured provider runtime. Use an eligible free allowance or review the linked plan before continuing; no guaranteed dollar cost.

Workflow design and an editorial reference output. No authenticated provider run is claimed. Product screens, available models and quotas can change.