Use it for a first draft
Ask it to explain, summarize, outline, compare, brainstorm, translate, organize, or help you practice.
Expanded Edition · 14 pages
The Companion Guide
A practical path from curious beginner to confident user—with prompt examples, privacy habits, verification steps, and a simple practice plan.
01 · Understand it
AI predicts useful responses. It does not know the way a person knows.
A conversational AI assistant analyzes patterns in large amounts of material and produces a likely response to your request.
Do not confuse fluency with truth
A polished answer may still be outdated, incomplete, biased, or invented.
Ask what is uncertain
Request assumptions, missing information, possible exceptions, and the claims that need verification.
Raise the standard with the stakes
Health, legal, financial, employment, and safety decisions require reliable sources and qualified people.
Your judgment matters.
AI can expand your options, but it does not become the final authority. You remain responsible for what you accept, share, and act on.
02 · Use it
A prompt is simply what you ask AI to do.
A useful first prompt explains the goal, safe background information, who will use the answer, the format, and any limits.
GoalContextReaderFormatLimitsCheck
Help me [GOAL]. The situation is [CONTEXT]. Write it for [READER]. Use [FORMAT]. Keep it [LIMITS]. Before answering, ask what important information is missing.
Try this complete example
Help me write a short, friendly text message to a coworker asking to switch shifts next Tuesday. Keep it under 80 words, make it respectful, and do not invent a reason. Before I send it, identify anything I should check.
“Reader” means the person who will read or use the answer—for example, a customer, coworker, student, or family member.
Learn something
Explain [TOPIC] in plain language for a complete beginner. Use one everyday example and finish with three check-your-understanding questions.
Compare choices
Compare [OPTION A] and [OPTION B] using [CRITERIA]. Separate known facts, assumptions, and questions I should investigate.
Improve the answer
Identify what may be outdated, uncertain, biased, or based on assumptions. Then revise the answer and add a verification list.
03 · Verify it
Verify before the answer becomes an action.
The more important the decision, the stronger your verification should be.
- Claim: What exact statement am I relying on?
- Source: Who published it, and are they qualified?
- Date: Is it current enough for this decision?
- Evidence: Can another reliable source confirm it?
- Context: What assumptions or exceptions change it?
- Decision: Who remains responsible for the final choice?
List the factual claims in your answer. For each one, tell me what kind of source could verify it, what may have changed, and what you are least certain about.
04 · Protect your information
Share the minimum necessary.
Privacy settings reduce risk; they do not make sensitive information invisible.
Usually safe
Public facts, invented examples, generic questions, and information you have permission to share.
Pause first
Internal work details, school records, unpublished plans, or information about another person.
Do not paste casually
Passwords, codes, account numbers, medical records, tax data, legal files, or confidential client material.
Your six-step settings check
- Open Privacy, Data Controls, or Activity settings. Look for a toggle such as “Improve the model,” “Help improve AI,” or similar. Switch it off if you do not want new chats used to train or improve the AI.
- Review saved memory, personalization, and chat history.
- Disconnect email, drives, calendars, and apps you do not need.
- Use temporary or private chat modes when appropriate.
- Use a unique password and two-step verification.
- Review active sessions, shared links, and old conversations.
ChatGPT data controlsGemini activity controlsClaude privacy guidanceMicrosoft Copilot privacy controlsGrok data controlsPerplexity data controls
What the training switch actually changes: Switching it off limits the use of new conversations for training or improving the AI. It does not make a conversation invisible or prevent the service from receiving and processing what you enter. Some information may also be kept temporarily for security, legal, or operational reasons. Continue to avoid entering sensitive information.
05 · Choose a tool
Start with the task, not the brand.
You can practice the same prompt in whichever assistant you already have. Use this comparison later when a different tool may better fit the task.
ChatGPT
Learning, writing, planning, analysis, and creating.
Gemini
General assistance with Google-connected options.
Copilot
Personal and Microsoft-connected assistance.
Claude
Writing, analysis, and working with long material.
Grok
Reasoning, voice, images, and current conversation.
Perplexity
Web research with visible source citations.
Features and policies change. Before connecting email, files, calendars, or contacts, review what the assistant may access, remember, or retain.
06 · Stay alert
AI can make false information look polished.
Scams, cloned voices, manipulated images, and misinformation often use urgency and emotional pressure.
Slow down
Urgency is a warning sign. Do not act from a message alone.
Change channels
Call a trusted number you already have—not one supplied in the message.
Use a family check
Create a private phrase or question for unusual money or emergency requests.
Verify identity
Familiar appearance or sound is no longer enough. Confirm through a separate route.
07 · Prepare
Build capability without surrendering agency.
AI will change tasks unevenly. Preparation means learning where it helps, where it fails, and what human responsibility still requires.
Map your tasks
Mark what is repetitive, creative, relational, confidential, regulated, or high stakes.
Practice one workflow
Use AI for a low-risk first draft, then measure time saved and errors found.
Strengthen human skills
Judgment, communication, trust, domain knowledge, ethics, and accountability remain valuable.
Keep evidence
Save strong prompts, verification steps, results, and lessons—not private data.
08 · Begin
A seven-day practice plan.
Spend about fifteen minutes a day. Use low-risk material and note what worked.
- Day 1Ask AI to explain a familiar topic. Correct what it gets wrong.
- Day 2Use the prompt formula for one small task.
- Day 3Ask it to improve its answer and state assumptions.
- Day 4Compare two options and verify one important claim.
- Day 5Review privacy, memory, connections, and security.
- Day 6Try a second assistant and compare the results.
- Day 7Choose one repeatable workflow and write your safety rules.