How Do I Write Good Prompts for AI Chatbots?

Here’s a secret that separates AI power users from everyone else: the model is rarely the difference — the prompt is. Give the same chatbot a vague request and a well-crafted prompt, and you’ll get work that’s unrecognisably better from the second. When people ask how do i write good prompts for ai chatbots, they’re really asking how to communicate with a brilliant, literal-minded alien that does exactly what you say rather than what you mean. Here’s the complete method.

The Core Formula: Context + Role + Task + Format + Constraints

Every great prompt contains five ingredients. Miss one and the output suffers predictably:

• Context: the background the AI needs. Who are you, what’s the situation, what has happened so far? AI can’t read your mind — supply the situation.
• Role: who should the AI be? “Act as an experienced employment solicitor” produces markedly different output from no role at all. Roles focus the model’s knowledge and tone.
• Task: the specific thing to do, stated as a verb. Not “about marketing” but “write three email subject lines for…”
• Format: how the answer should look — bullet points, table, 200 words, step-by-step, with headings. AI defaults to essays; tell it what you actually want.
• Constraints: length limits, tone, what to avoid, who the audience is. “Explain like I’m a smart 15-year-old” beats “explain simply” every time.

Before and After: See the Difference

Weak prompt: “Write about email marketing.”
Result: a generic 500-word essay you could have Googled.

Strong prompt: “Act as an email marketing specialist for UK small businesses. My bakery in Leeds has 800 newsletter subscribers and low open rates (18%). Give me 5 specific tactics to improve open rates, each with a concrete example I could implement this week. Format as a numbered list with bold headings. Practical tone, no jargon.”
Result: tailored, actionable, immediately usable advice.

The strong prompt took 30 seconds longer to write and produced 10× the value. That ratio holds across everything — it’s the highest-leverage half-minute in modern knowledge work.

Advanced Techniques That Actually Help

Give examples (few-shot prompting)

Show the AI what “good” looks like: “Here are two product descriptions I like: [examples]. Write three more in the same style for these products: [list].” Examples calibrate tone, length and style faster than any description.

Ask it to ask you questions

“I need to write a difficult email to a client. Before drafting, ask me the 5 questions you need answered to get the tone right.” This flips the dynamic — the AI extracts the context you’d have forgotten to include.

Request reasoning

“Explain your reasoning step by step before giving the final recommendation.” For complex decisions, seeing the working lets you spot flawed assumptions — and the final answer is usually better for it.

Iterate, don’t restart

The first answer is a draft. Follow up: “Make it shorter,” “more formal,” “give me the contrarian view,” “what am I missing?” Three rounds of refinement routinely beat one perfect prompt — because you often don’t know what you want until you see a first attempt.

Set the audience explicitly

“Explain this for a board of non-technical directors” vs “explain this for a junior developer” — same facts, completely different outputs. Audience is the most underused prompt ingredient.

Common Prompting Mistakes

• Asking for facts without verification: AI invents confidently. For anything factual, add “flag anything you’re uncertain about” — and verify independently.
• One giant prompt for a complex task: break big jobs into steps. Outline first, then expand section by section. Quality compounds.
• Accepting the first draft: see above — iterate.
• No constraints: without length and format guidance, you get the AI’s defaults, which are rarely what you wanted.
• Forgetting context across sessions: each new chat starts fresh. For ongoing projects, paste a brief recap at the start.

Prompts for Common Tasks (Steal These)

• Email: see our full guide to using AI for better emails.
• Learning: “Explain [concept] to me as if I’m a clever 14-year-old. Use one analogy, then quiz me with 3 questions.”
• Decisions: “I’m deciding between [A] and [B]. Act as a neutral adviser. List the strongest arguments for each, the key uncertainties, and what additional information would most change the decision.”
• Writing feedback: “Review this draft for clarity and persuasiveness. List the 3 weakest points and suggest specific fixes. Be direct.”
• Planning: “Help me plan [project] over 4 weeks. Break it into weekly milestones, flag dependencies, and identify the biggest risks.”

Prompting for Images and Other Media

Everything in this guide transfers to image generation — with a few medium-specific twists. Structure image prompts as: subject + style + composition + lighting + mood + details. “A lighthouse” is a wish; “a solitary lighthouse on rugged Cornish cliffs at dusk, dramatic storm clouds, oil painting style, moody and atmospheric, wide landscape composition” is a commission. Style references are the image prompter’s superpower: “in the style of a 1950s travel poster,” “cinematic still,” “editorial photography” — each invokes a rich visual tradition the model understands.

Negative prompts (what to exclude: “no text, no watermark, no extra fingers”) solve the most common defects preemptively. Aspect ratio and composition matter enormously — specify portrait/landscape/square and where the subject sits. And the iteration principle intensifies: professional AI artists generate dozens of variations, curating ruthlessly. The prompt gets you 70% there; selection and refinement deliver the final 30%. Same for video and audio generation: describe the shot, the movement, the mood — then curate. Across every medium, the meta-skill is identical: clear vision, specific language, relentless iteration. Master prompting in text and you’ve mastered the grammar of every generative medium — the vocabulary just changes.

Related reading: put your prompting skills to work with our roundup of AI tools for writing blog posts.

For the technical background, see this introduction to machine learning.

Practical Takeaway

Writing good prompts means supplying context, role, task, format and constraints — the five ingredients that turn generic output into tailored results. Show examples of what good looks like, iterate rather than accepting first drafts, break complex jobs into steps, and always verify factual claims. Master this and every AI tool you touch — chatbots, image generators, coding assistants — becomes dramatically more useful. Prompting isn’t a party trick; it’s the literacy of the AI age, and thirty seconds of extra thought routinely returns ten times the value.