Same prompt. Three models. Three formats.
Each AI model reads your prompt differently. We format it for the one you're actually using.
Your prompt, formatted for:
<task>
Write a marketing email promoting FocusFlow, a new productivity app for freelancers. The email should drive sign-ups for a 14-day free trial.
</task>
<context>
FocusFlow brand voice guide (excerpt for marketing copy):
We talk to freelancers like a smart friend who happens to know a lot about productivity tools, not like a SaaS company trying to close a deal. That means contractions, short sentences, and zero jargon like "leverage," "synergy," or "seamless."
Tone: warm, a little wry, never salesy. We're allowed to be funny about the chaos of freelance life (the seventeen open tabs, the client who emails at 11pm) because our users live it too. We are not allowed to be funny about money, deadlines they've missed, or anything that could read as mocking someone's actual stress.
What we never say: "revolutionary," "game-changing," "10x," "unlock your potential," or anything that sounds like a keynote slide. If a sentence could appear in a startup's Series A deck, cut it.
What we always do: lead with the specific problem before the feature. Freelancers don't care that FocusFlow has "AI-powered context switching detection" β they care that it stops them from losing 20 minutes every time a client Slack message interrupts deep work. Name the pain first, in their words, then the fix.
Formatting preferences: short paragraphs (2-3 sentences max), one idea per paragraph, and when giving instructions or steps, prefer a simple numbered or bulleted list over a dense paragraph. Subject lines should feel like they're from a person, not a newsletter β no emoji, no "π," no ALL CAPS urgency.
Call-to-action language: "Try it free for 14 days" beats "Start your free trial" beats "Sign up now." We want it to sound like an invitation, not a funnel step. Never use countdown-timer urgency ("Offer ends soon!") β our users can smell fake scarcity and it actively hurts trust with this audience.
A note on humor: self-deprecating is good, punching down is not. We can joke about our own onboarding flow being clunky in v1. We do not joke about other productivity apps by name β freelancers talk to each other, and dunking on a competitor reads as insecure, not confident.
</context>
<audience>
Freelancers and independent contractors who juggle multiple client projects and struggle with context-switching.
</audience>
<tone>
Casual and encouraging β like a friend who found something useful, not a corporation selling something.
</tone>
<format>
Under 200 words. One clear call-to-action near the end. Subject line included.
</format>
<constraints>
Do not mention competitor products by name. Do not use the word "revolutionary." No ALL CAPS.
</constraints>
<success_criteria>
The reader clicks the free trial link. Secondary: they forward it to a freelancer friend.
</success_criteria>Gemini wants your context first.
Googleβs documentation says to put instructions after long context and open with "Based on the preceding informationβ¦" β the opposite of what most people do.
Source βChatGPT reads Markdown better than XML.
GPT-5.x is trained on more Markdown than XML. Structured headers give it cleaner signal than angle brackets.
Source βClaude is the opposite.
Claudeβs own documentation recommends XML-style tags for section boundaries. Same information, different packaging.
Source βKnow what's missing before you send it
Most advice about writing better prompts is either vague or unproven. We built a completeness check instead β it tells you specifically what's missing from your prompt, based on the one thing research actually shows matters: whether you've said enough.
Completeness: 3/7
2/7 β 5/7 β nice, that's more complete.
Format
No length or structure is specified β "a blog post" could mean 300 words or 3,000.
Fix: Specify a target length and structure (e.g. "600β800 words, intro plus three sections").
This isn't a quality score β research shows prompt effectiveness varies too much by task and model to promise that. It's a completeness check: does your prompt say enough for an AI to work with? That part is measurable, and it's exactly what our guided questions help you fix.
PromptMe Pro β $5/month
Everything in the free builder, plus:
- Copy your prompt formatted for Claude, ChatGPT, or Gemini
- Each format follows that model's own documentation
- New model formats added as guidance changes
- A completeness check with specific, actionable fixes β not just a score
The builder itself is always free. Build unlimited prompts, get a structured result you can paste anywhere.
No credit card charge for 7 days. Cancel anytime.
By starting your trial you agree to our Terms of Service and Privacy Policy.
Just want the free builder? Build a prompt β it's free β
