Blog

How to Write Better ChatGPT Prompts

ChatGPT is capable of producing research summaries, polished drafts, detailed plans, and analysis — but only if the prompt is specific enough to guide it there. This guide covers the techniques that make the biggest practical difference in ChatGPT output quality.

What makes ChatGPT prompts different

ChatGPT is a general-purpose model, which means it can handle an enormous range of tasks — but that flexibility is also why vague prompts produce mediocre results. When you ask ChatGPT a broad question, it generates a statistically likely response. That response is usually safe, generic, and organized — but it rarely contains the specific insight, angle, or format you actually need.

The fix is not to make prompts longer. It's to make them more specific. A 20-word prompt with a clear task, defined audience, and requested format will outperform a 200-word prompt full of vague intentions. Specificity is the lever, not length.

The "Act as" framing technique

One of the most effective ChatGPT techniques is to start your prompt with a role definition: "Act as a [specific expert] with [relevant experience]." This frames the model's response within a particular perspective, knowledge base, and communication style.

Instead of asking ChatGPT to "review my business plan," ask it to "act as a venture capital analyst with experience evaluating early-stage SaaS companies and review this business plan for the top 5 risks an investor would flag." The role changes what the model pays attention to and how it structures its response.

The most useful roles are specific rather than generic. "Act as a senior copywriter" gives more direction than "act as an expert." "Act as a hiring manager at a tech startup reviewing resumes for a UX role" gives far more direction than "act as a hiring manager."

Using constraints to sharpen output

Constraints are the most underused element in most prompts. A constraint tells ChatGPT what to avoid, what limits to respect, or what standards to meet. Without constraints, the model fills space with what it knows — which often includes filler, hedging, and over-explanation.

Useful constraints include: word count limits ("under 150 words"), format restrictions ("no bullet points — write in paragraphs"), style rules ("don't use the word 'leverage' or any corporate jargon"), and scope limits ("focus only on organic tactics, not paid advertising"). Each constraint reduces the solution space and forces more focused output.

Practical prompt examples by use case

Writing / editing

Act as a professional editor. Rewrite this paragraph to be 30% shorter while preserving all key information. Remove passive voice, cut filler phrases, and make the opening sentence stronger. [paste paragraph]

Business planning

Act as a business strategist. Create a 90-day action plan for a local service business that wants to increase leads by 30%. Prioritize tactics by impact-to-effort ratio. Return as a numbered list with weekly milestones.

SEO content

Act as an SEO content strategist. Given the keyword 'best email marketing tools for small business,' create a content brief that includes: search intent, recommended title tag, H2 structure, word count recommendation, and 3 internal linking opportunities.

Job searching

Act as a hiring manager at a mid-size tech company. Review this resume summary and tell me: (a) what impression it creates in 6 seconds, (b) what's missing that would make it stronger, (c) a rewritten version. [paste summary]

Customer emails

Act as a customer success manager. Write a reply to this customer who is frustrated about a billing error. Be empathetic, take ownership, explain clearly, and give a specific resolution. Keep it under 120 words. [paste customer message]

Summarization

Act as a research analyst. Summarize this 800-word article into: (1) a 3-sentence executive summary, (2) the 5 most important points, (3) 2 questions this article doesn't answer. [paste article]

The iteration technique: getting better second drafts

Most people treat the first ChatGPT response as the final answer. Professionals treat it as a first draft. The fastest way to improve output is to use the initial response as a foundation and then give specific improvement instructions.

After the first response, follow up with targeted edits: "The third bullet is too vague — make it specific with an example." "The intro assumes too much background — rewrite it for someone who has never heard of this topic." "Add a section addressing the most common objection someone would have." Each follow-up narrows the gap between what you got and what you need.

Common ChatGPT prompt mistakes

Related tools and resources