Learn to talk to AI.
Get answers that actually help.
Six simple principles, side-by-side prompt examples, a library you can copy today, and a plain-English comparison of ChatGPT, Claude, Gemini, and the rest, free vs paid.
The six ideas that change everything about prompting.
AI isn't magic. Think of it as a very well-read intern. Whether you get a brilliant answer or a useless one mostly comes down to how you ask. Internalize these six and you're already ahead of 90% of people using AI.
Be specific
Vague questions get vague answers. Spell out the topic, audience, length, and what a good answer looks like.
Give context
Tell the model who you are, what you're working on, and what you've already tried. Context beats cleverness.
Show, don't just tell
Paste an example of the output you want. One or two samples dramatically improves quality.
Assign a role
“Act as a senior copy editor…” nudges tone, vocabulary, and depth in a single sentence.
Constrain the output
Word count, bullet points, tone, format (JSON, table, markdown). Constraints make results usable.
Iterate, don't restart
Treat it as a conversation. Say “shorter,” “more technical,” “rewrite the 2nd paragraph” instead of starting over.
Same task. Two very different prompts.
The vague version is what most people type. The sharp version is what gets results. Read both, and the pattern becomes obvious fast.
Writing an email
Write an email about the meeting.
Write a 120-word email to my team recapping today's product meeting. Tone: friendly but professional. Include: (1) the decision to ship v2 on May 5, (2) that QA owns the launch checklist, (3) a request for blockers by Friday. Sign it “Sam.”
Why it works: The good version names the audience, length, tone, the three points to cover, and the sign-off. Nothing to guess.
Drafting a client proposal
Write me a proposal for a new client.
Draft a 1-page proposal for a prospective client, Acme Co., a 50-person logistics company. We're proposing a $12k, 6-week engagement to audit and rebuild their onboarding workflow. Tone: confident, not salesy. Sections: the problem we heard from them, our approach in 3 phases, deliverables, timeline, and pricing. Leave a `[NAME]` placeholder for their main contact.
Why it works: Spells out the audience, deal size, scope, structure, tone, and even where to leave placeholders, so the draft is usable instead of a blank-slate guess.
Preparing for a meeting
Help me prepare for my meeting tomorrow.
I have a 45-minute Q3 planning meeting tomorrow with my VP and two peers. My goal: get sign-off on hiring two more analysts. Give me (1) a 3-bullet opening that frames the ask in business terms, (2) the 5 questions they'll most likely push back with, with a one-line response to each, and (3) the single strongest data point I should lead with.
Why it works: Names the meeting type, attendees, time budget, and your real goal, so the prep is tactical instead of generic.
Turning a meeting into action items
Summarize this meeting transcript.
From the transcript below, extract: (1) every decision made, with who decided it, (2) every action item as `Owner / Task / Due date` (mark “unassigned” if unclear), (3) any open questions that didn't get resolved. Skip small talk. Format as three markdown lists. Transcript: <paste>
Why it works: Defines the exact output the team can drop into a project tracker, instead of a wall-of-text recap nobody reads.
Summarizing
Summarize this article.
Summarize the article below for a busy executive. Output: a 2-sentence TL;DR, then 5 bullets of key claims, then 2 bullets of what the author leaves unanswered. Keep it under 180 words. Article: <paste>
Why it works: Defines the reader, the structure, the length, and asks the model to surface gaps instead of just regurgitating.
Copy these. Paste them into any AI.
Battle-tested prompts you can adapt in under a minute. Anywhere you see <paste>, drop in your own text.
Sharpen any paragraph
Rewrite the paragraph below to be clearer and 30% shorter, without losing any facts. Use plain English. Then list the 3 biggest changes you made and why. Paragraph: <paste>
Tone shift
Rewrite this message three times: (1) warm and casual, (2) professional and concise, (3) direct and slightly firm. Keep every version under 80 words. Message: <paste>
Explain like I'm 12
Explain <topic> to a curious 12-year-old using one everyday analogy the whole way through. End with 3 questions I could ask to test whether I really understood it.
Study plan
I want to learn <skill> in 30 days, studying ~45 minutes a day. Build me a week-by-week plan with concrete daily tasks, free resources, and a small project I can show at the end of each week.
Every major AI model, in plain English.
What each one is actually good at, where it falls short, and what the free tier gets you vs. paying.
ChatGPT
The generalist. Big ecosystem, tons of features.
- Everyday writing & brainstorming
- Image generation (DALL·E)
- Custom GPTs & plugins
- Voice mode & file uploads
- Free tier has tighter limits during peak
- Can be overly cautious on edge-case asks
GPT-5 mini with limits, basic voice, limited image generation, light file uploads.
Plus ($20): full GPT-5.4 Thinking, Deep Research, Sora, Codex, Agent Mode. Pro ($200): GPT-5.4 Pro, ~680-page context, 250 deep-research runs/mo.
Claude
The thoughtful writer and coder. Long, careful answers.
- Long-form writing & editing
- Reading big documents (1M-token context)
- Coding and multi-step reasoning
- Nuanced, steerable tone
- No native image generation
- Web search limited vs. competitors
Claude Sonnet access with daily message caps, file uploads, Projects (limited).
Pro ($20): Claude Code in the terminal, unlimited Projects, Workspace integration, extended reasoning. Max ($100/$200): 5× or 20× the usage and priority access.
Gemini
Deeply wired into Google. Great at research and multimodal.
- Research with live Google Search grounding
- Working inside Gmail, Docs, Sheets
- Huge context windows & video understanding
- Free multimodal generation
- Writing voice can feel generic
- Occasional over-refusals
Gemini 2.5 Flash unlimited, Pro with daily cap, image gen, basic Deep Research.
AI Pro ($19.99): Gemini 3 Pro, 2TB storage, Workspace add-ins. AI Ultra ($249.99): Gemini 3.1 Pro, Deep Think, Veo 3.1, 25k AI credits/mo.
Perplexity
An answer engine, not a chatbot. Cites every claim.
- Research questions with sources
- Comparing products, prices, facts
- Quick “what is the latest on…”
- Not a great creative writer
- Answers can be shallow without Pro
Quick searches, limited Pro searches per day.
Pro ($20): 300+ Pro searches/day, choose model (GPT, Claude, Gemini), file analysis, Spaces. Max ($200): unlimited use, Labs, Comet browser, Perplexity Computer.
Which AI for which job?
If you only remember one thing from this page, let it be this table.
- Write / edit long documentsClaudeHandles book-length context and keeps voice consistent.
- Code & debuggingClaude or ChatGPTBoth top-tier; Claude is stronger for multi-file reasoning.
- Research with citationsPerplexityBuilt specifically to ground answers in real sources.
- Latest news / live webGeminiNative Google Search grounding gives fresh, sourced results.
- Image generationChatGPT or GeminiIncluded in the free tiers; quality is consistently strong.
- Inside Gmail / Docs / SheetsGeminiNative Workspace access reads and writes your files.
- Voice conversationsChatGPTAdvanced Voice Mode feels closest to talking to a person.
- Comparing vendors or productsPerplexitySide-by-side answers with the receipts attached.
- Free everyday use at workGemini Flash or ChatGPT freeBoth have generous free tiers that cover most daily tasks.