On the way to become great at clipping

https://www.instagram.com/reel/DbUKc3atZ74/?igsh=aTd3OW9jZHhub3Nh

Most people get bad output from ChatGPT/Gemini not because the model is weak — it's because the prompt is lazy.

Three quick fixes that instantly improve any AI output:

1. Give it a role ("You are a senior copywriter for a DTC brand...") instead of a task alone.

2. Force structure — ask for output in a specific format (bullets, table, JSON) so you're not editing rambling paragraphs.

3. Chain it — generate a draft, then run a second prompt asking the model to critique and rewrite its own output. This alone doubles quality.

I build ready-to-use prompt packs for creators/marketers over at PromptForge because most people don't want to learn prompt engineering — they just want prompts that already work. Just launched AI Prompt Library: a tool with 1,000+ plug-and-play prompts for content, marketing, and viral hooks — one-time $25, no subscription.

Happy to answer prompt engineering questions here if useful.

Been building AI prompt packs for creators for a while now, and the biggest thing I've noticed: most people write prompts that ask for a caption. That's the wrong ask.

The better move is asking the model for a structure, not a sentence. Something like:

"Give me 5 hook variations for [topic] using: a controversial claim, a relatable pain point, a number-driven stat, a question, and a mini-story opener. Each under 12 words."

You get 5 angles, not 5 rewordings of the same idea. Then you pick the one that fits your voice.

I just launched PromptForge — a weekly-updated vault of these exact kinds of structured prompts for hooks, captions, and content ideas, built specifically for creators and marketers who post often and don't have time to reinvent the wheel every day.

If you want a look: whop.com/promptforge-f219

Happy to answer questions on prompt structuring in the comments 👇