A growing library of copy-ready prompt templates for writing, coding, marketing, research, study and image generation. Each one uses the structure that consistently produces high-quality output from ChatGPT, Claude, Gemini and Midjourney.
Across writing, coding, marketing, research, study, image.
Updated when models change and based on real usage data.
No license, no signup. Use in personal or commercial work.
Works in ChatGPT, Claude, Gemini, Perplexity, Mistral, Llama.
Copy any template, replace the [variables], and run it. Every template uses the same six-part structure (role, audience, task, format, constraints, examples) that consistently lifts answer quality.
Use after outlining — fills draft fast.
You are a senior writer for [publication]. Write a [word count] article on [topic] for [audience]. Tone: [voice]. Structure: H1, 4 H2 sections, conclusion. Lead with a concrete pain point. Avoid: clichés, hedging, listicle filler.
Skipping known context = much tighter output.
Summarize the text below for [audience]. Format: [N] bullets, each ≤[N] words, ordered by [importance / chronology]. Skip context the reader already has. Surface only [decisions / facts / actions].
Rationale-per-change makes PR comments easy.
You are a senior [language] reviewer. Review the function below for runtime errors, type-safety issues, framework anti-patterns. Return corrected code + 1-line rationale per change. No style nits.
Ranked hypotheses → fewer wrong fixes.
Given the error + code below, produce: (1) most likely root cause, (2) 2 alternative hypotheses ranked, (3) minimal repro, (4) suggested fix with diff. Stack: [stack].
Variants for split testing.
Write [N] [channel] ad variants for [product]. Audience: [persona]. Constraints: [char limits]. Lead with a concrete outcome, end with a verb-led CTA. No exclamation marks.
Strict char counts force clarity.
Write a landing page hero for [product] targeting [persona]. Output: H1 (≤9 words), subhead (≤22 words), 3 outcome-led bullets, 4-word CTA. Voice: [voice].
Conflict flag prevents one-sided summaries.
You are a research assistant. Summarize the key findings on [topic] from the sources below. Output: 5 findings, each with citation, confidence (high/med/low), and 1-line implication. Flag conflicting evidence.
Forces the model to commit to placements.
You are a strategy consultant. Map the competitive landscape for [market]. Output: a 5×5 matrix with axes [axis X] and [axis Y], placing 8-12 competitors. For each competitor: positioning sentence + their strongest moat. Then surface the 2 most underserved cells.
Personalized depth + active recall.
Explain [concept] to a [year] student who already understands [prerequisite]. Use one everyday-life analogy, then 2 worked examples of increasing difficulty. End with 3 self-check questions.
Tutor mode, not answer-machine mode.
Read my work below. Do NOT rewrite it. Ask me [N] Socratic questions designed to expose weak arguments, missing evidence or unclear definitions. Order by importance.
Subject + style + composition + lighting + lens.
[Subject in 1 sentence]. Style: [style]. Composition: [framing]. Lighting: [lighting + mood]. Lens: [focal length + aperture] (for photoreal). --ar [ratio] --v 6
Low stylize keeps the product accurate.
Product photography of [product] on a [surface]. Lighting: [3-point / softbox / natural]. Lens: 50mm macro, f/2.8. Background: [neutral / contextual]. Style: [editorial / catalogue / lifestyle]. --ar 3:2 --v 6 --style raw --stylize 100
Every template here includes role, audience, task, format, constraints, examples — the six dimensions that drive answer quality.
A filled template lands a strong first answer ~3-5x more often than a blank prompt.
Shared templates mean shared output quality. Especially valuable for marketing, content and engineering teams.
Templates can be chained: outline template → draft template → critique template. Each step adds quality.
These templates work in ChatGPT, Claude, Gemini, Mistral and Llama. The structure transfers across LLMs.
Paste a filled template into AI Prompt Fixer to adapt audience, voice and constraints to your exact case.
Copy any template, replace the [variables], then paste it into the free AI Prompt Fixer to adapt it to your exact audience and voice.
