Text guides
Getting useful AI help
Give an AI request enough context and ask for a check that matters. Keep facts, assumptions and unfinished actions visible in the reply.
12 practical tasks. Each guide includes a sample request and a way to check the answer.
- Turn messy notes into a clear AI task briefSort your notes into five parts: the result you want, the facts you supply, the limits that apply, the format you need back, and what is still undecided. Put each fact on its own line so nothing hides inside a paragraph.
- Ask an AI assistant to identify missing information firstBefore asking for the finished answer, ask what the assistant would need to know to give a reliable one. Tell it to separate required unknowns from optional ones and to draft nothing yet. Then supply the facts and request the real output.
- Set source requirements for an AI research requestDecide before the request what counts as a good source, how recent it must be, and what each citation must include. Put those rules in the question and ask the assistant to mark any claim it cannot support, so weak spots are easy to find.
- Set comparison criteria before asking for a recommendationWrite the conditions an option must meet, the preferences that matter and anything that rules it out. Ask for a comparison against that list, with unknowns shown clearly. Keeping the criteria visible lets you see why a recommendation follows and whether the same standard was applied to every option.
- Ask one AI agent to draft and another to checkA useful review needs a question the draft could fail. For a client proposal, that might be whether every promised deliverable appears in the agreed brief. Give the reviewer the original requirements so it can check the draft against something concrete.
- Ask for an assumption check on an AI-generated planTake each step of the plan and ask what must be true for it to work. Sort those conditions into facts you confirmed and assumptions you did not. Then note what happens if an assumption is wrong and how you would check it.
- Correct an AI answer without restarting the whole requestName the wrong detail, give the correct fact with its source, and list the parts of the answer that depend on it. Ask for a revision of only those parts. This keeps the work that was right and shows where the change landed.
- Ask for an answer that is usable on a phoneTell the AI you will read the answer on a phone and request a format that fits the task. Start with a short answer, then a small checklist and the next action. Include an unresolved question only when one remains, and keep conditions beside the steps they affect.
- Separate supplied facts from AI assumptionsAsk the AI to tag every important statement as supplied by you, inferred, or unknown, and to name what would verify each inference. That shows which claims rest on your material and which are the AI filling gaps.
- Check an AI result before sending or buying anythingA draft, recommendation or plan from an AI is text, not a completed action. Before you send or pay for anything, confirm each exact detail against its source, confirm who has authority to approve, and record that the action has not yet happened.
- Summarize a long AI conversation for the next taskAsk the AI for a short handoff note that carries forward only what the next task needs: the goal, verified facts, your decisions, rejected options and the next request. Leave the rest of the conversation behind.
- Ask a useful follow-up when AI agents disagreePut the competing claims beside their evidence and assumptions. Check whether the agents answered the same question, then identify the missing information or differing criteria behind the disagreement. Some disputes need several facts or a decision about priorities; agreement among agents does not settle them.
Related topics
- Research and sources
Start with a question, keep a record of evidence and show where sources disagree. Ask for help organizing research without treating a generated answer as proof.
- Content publishing
Turn your knowledge and source material into useful pages. Keep claims traceable, give readers a clear answer and make the next edit obvious.
- Product planning
Turn an idea into a small, testable piece of work. State what users need, what the first version includes and how to check it.