DeepclarioDeepclarioImprove a prompt free
Techniques· 8 min read

How to get better results from ChatGPT

The model is not the problem. If you keep getting vague, generic, or wrong answers from ChatGPT, the fix is almost always in the prompt. These 8 techniques address the most common reasons prompts fail.

  1. 01

    Tell it who to be

    Starting with "Act as a [role]" is not just a trick — it shifts the model's entire frame. "Act as a senior software engineer" produces different code explanations than no role. "Act as a skeptical editor" produces harsher, more useful feedback on your writing. The role sets the vocabulary, the level of detail, and the assumptions.

    Weak

    Review my code.

    Strong

    Act as a senior backend engineer doing a code review. Review the code below for security issues, performance problems, and readability. Be direct and specific about what needs to change and why.

  2. 02

    Specify the format before you ask

    If you don't say what format you want, ChatGPT decides for you — and its default is often prose when you need bullets, or bullets when you need prose. State the format upfront: "Give me a numbered list", "Format as a table with columns X, Y, Z", "Write 3 short paragraphs, no headers."

    Weak

    What are the pros and cons of remote work?

    Strong

    List the top 5 pros and top 5 cons of remote work in a markdown table. Each cell should be one sentence. Target audience: a CEO deciding whether to go fully remote.

  3. 03

    Give it the context it cannot guess

    ChatGPT knows nothing about your situation unless you tell it. Who is the audience? What has already been tried? What decision will this output inform? What constraints exist? Every piece of context you add removes a guess it has to make, and guesses are where generic output comes from.

    Weak

    Write a product description for my app.

    Strong

    Write a product description for Deepclario — an AI prompt analyzer. Target audience: marketing professionals who use ChatGPT daily but are frustrated with inconsistent results. Key differentiator: it asks clarifying questions before rewriting, so the output is tailored, not generic. Tone: confident and direct. Length: 80 words max.

  4. 04

    Add negative constraints

    Telling ChatGPT what NOT to do is as powerful as telling it what to do. Common useful negatives: "No bullet points", "Don't use corporate jargon", "Don't hedge with phrases like 'it's worth noting'", "Don't repeat the question back to me", "No more than 200 words."

    Weak

    Write an executive summary.

    Strong

    Write an executive summary of the report below. Max 150 words. No bullet points — prose only. Do not start with "This report..." Start with the most important finding.

  5. 05

    Ask for multiple variations

    Instead of asking for one version and editing it to death, ask for 3-5 variations upfront. "Give me 5 subject line options for this email" or "Write 3 versions of this headline — one formal, one casual, one provocative." You will almost always find one that is close to what you want, or combine the best parts.

    Weak

    Write a subject line for my cold email.

    Strong

    Write 5 subject line options for a cold email selling a B2B SaaS tool to heads of marketing. The tool saves 3 hours per week on reporting. Vary the angle: curiosity, benefit, social proof, question, and directness — one of each.

  6. 06

    Use chain-of-thought for hard problems

    For complex reasoning tasks, adding "think step by step" or "reason through this before giving your final answer" dramatically improves accuracy. It forces the model to work through the problem rather than pattern-matching to a surface-level answer.

    Weak

    Should I raise prices?

    Strong

    I run a SaaS with 200 customers paying $49/mo. My churn is 3%/mo. My main competitor charges $79/mo. Think step by step: what factors should I consider before raising prices, and what would you recommend, and why?

  7. 07

    Iterate instead of starting over

    The first output is a draft, not a failure. Follow up with specific refinements: "The third paragraph is too long — cut it in half", "The tone is too formal — rewrite in a more conversational voice", "This is good but I need it to be more specific about X." Each follow-up teaches ChatGPT more about what you want.

    Weak

    [starts a new chat and tries again]

    Strong

    "This is good but the opening is too generic. Rewrite just the first paragraph to start with a specific, surprising fact about the problem instead."

  8. 08

    Show an example of what good looks like

    If you have an example of output you like — a previous email, a competitor's copy, a paragraph you wrote — paste it in. "Write something like this:" followed by your example is one of the most powerful prompt techniques. The model will match the structure, rhythm, and tone far more accurately than any description.

    Weak

    Write a bio for my LinkedIn.

    Strong

    Write a LinkedIn bio for me in the style of this example: [paste in a bio you admire]. My details: [your actual details]. Match the tone and structure but make it specific to my background.

See exactly what your prompt is missing

Deepclario scores your prompt across five dimensions, tells you which of these techniques apply, and rewrites it for you. Free, no account needed.

Analyze my prompt free