You are a social media copywriter who has written hooks for posts with millions of views on [PLATFORM: Instagram / TikTok / LinkedIn / Twitter]. Generate 10 scroll-stopping hooks for a post about [TOPIC] aimed at [TARGET AUDIENCE].
Each hook must:
- Be under 12 words
- Create curiosity, tension, or contrarian surprise
- Avoid clickbait — deliver on what it promises
- Match the native voice of [PLATFORM]
- Use a different angle (statistic, question, mistake, story-open, before/after, listicle tease, hot take, callout, confession, prediction)
Label each hook with its angle. Then rank the top 3 most likely to perform and explain why.
Why this works
Forcing 10 different angles prevents the model from giving you 10 versions of the same hook. The ranking step gives you a starting point so you don't have to evaluate them all yourself.
Write a [PLATFORM] caption for a post about [POST TOPIC]. The visual is [DESCRIBE PHOTO/VIDEO/CAROUSEL].
Structure:
- Line 1: a hook under 10 words that earns the "see more" tap
- Lines 2–5: the value, story, or insight (short paragraphs, one idea each)
- Final line: one clear CTA — [COMMENT / SHARE / SAVE / DM / LINK IN BIO]
Voice: [casual / authoritative / vulnerable / playful]. No hashtags inside the body. End with a question that invites a real reply, not "thoughts?".
Why this works
Locking in the first-line hook plus a single CTA prevents the model's instinct to pile on three competing calls-to-action. Banning "thoughts?" forces a specific question that actually drives comments.
Build a hashtag strategy for a [PLATFORM] account in the [NICHE] space targeting [AUDIENCE].
Give me 30 hashtags split into 3 tiers:
- Tier 1 (5 small/niche, under 50k posts): low competition, high relevance
- Tier 2 (15 mid-size, 50k–500k posts): the workhorse tier
- Tier 3 (10 large/branded, 500k+ posts): broad reach, lower hit rate
For each tag, add a short reason. Then list 5 hashtags I should AVOID (banned, spammy, or so generic they hurt reach) and explain why. Finish with a rotation plan for how to vary tags across posts.
Why this works
Tiering by post volume mirrors how growth-focused accounts actually mix hashtags — niche tags do the discovery work, big tags add ceiling. The "avoid" list catches the spammy defaults the model would otherwise suggest.
Build a 30-day [PLATFORM] content calendar for [BRAND/CREATOR NAME] in the [NICHE] space. My goals are [GOAL 1, GOAL 2 — e.g. follower growth, email signups]. My audience is [AUDIENCE].
Output as a table with columns: Day | Format (Reel / Carousel / Static / Story / Live) | Hook | Topic | CTA | Pillar.
Use 4 content pillars: [PILLAR 1], [PILLAR 2], [PILLAR 3], [PILLAR 4]. Distribute formats so I post 3–4x per week, not daily. Include 2 "experiment" posts that test a new format. End with one weekly theme summary.
Why this works
Pillars stop the model from drifting into 30 random topics. The 3–4x cadence and "experiment" slots match what's actually sustainable — daily posting calendars usually get abandoned in week two.
Here is a viral [PLATFORM] post that did [METRIC — e.g. 2M views, 80k likes]:
[PASTE FULL POST TEXT / DESCRIBE VIDEO BEAT-BY-BEAT]
Reverse-engineer why it worked. Break down:
- The hook (what tension/curiosity it created)
- The structural pattern (problem→agitate→solve, listicle, story, contrarian take, etc.)
- The pacing (where attention drops are prevented)
- The emotional payoff in the final line
- The CTA — explicit or implicit
Then write 3 original posts of mine on [MY TOPIC] that copy the STRUCTURE but not the wording. Label each with the structural element it borrows.
Why this works
Copying STRUCTURE not wording is how every great copywriter learns. Forcing the model to label which element it's borrowing teaches you the patterns instead of just handing you knockoff content.
I'm replying to comments on a [PLATFORM] post about [POST TOPIC]. My account voice is [VOICE — e.g. friendly expert, dry/witty, supportive].
Here are the comments to reply to:
[PASTE COMMENTS, ONE PER LINE]
For each, write a reply that:
- Sounds like a human, not a brand auto-reply
- Is 1–3 short sentences
- Either adds value, asks a follow-up, or makes the commenter feel seen
- Uses NO emojis unless the original comment used one
Flag any comments I should NOT engage with (trolls, spam, bait) and explain why.
Why this works
Reply length and emoji rules stop the model from generating those obviously-AI "Thanks so much for sharing!" messages. The "do not engage" flag saves you from feeding trolls or arguing with bait accounts.
Write a 7-frame Instagram or TikTok Story script for a launch sequence. Each frame includes the on-screen text, voice-over line, sticker call-to-action, and timing. Builds urgency without feeling pushy. Includes a swipe-up final frame and a recovery frame for people who tap away early...
Turn any topic into a 10-slide carousel with cover hook, 8 value slides, and a save-worthy summary slide. Each slide gets a headline, body line, and visual direction. Optimized for Instagram and LinkedIn save rates...
Generate 50 engagement questions sorted by intent: identity, opinion, story, vote, advice. Designed to actually drive comments — not the "what did you think?" filler that gets ignored. Includes platform-specific tweaks for Reels, LinkedIn, and Twitter polls...
Craft 5 versions of a profile bio for Instagram, TikTok, LinkedIn, and Twitter — each tuned to that platform's character limits, search behavior, and conversion psychology. Includes link-in-bio CTA testing variants and keyword placement for discovery...