The 6-Rule Carousel System: Structure, Cover Lines, and Captions With AI

The problem with most carousels

Most carousels fail not because the content is bad, but because they're structured like a blog post that someone turned sideways. Long intro slide. Vague topic label. Content that only makes sense read in order. No clear reason to save.

Six rules fix this. They govern how a carousel should be built — from the first word of the cover to the last line of the caption.

Rule 1: The cover has one job — stop the scroll

Eight words maximum. A promise or a number, never a topic label.

"The 5 ways to improve your focus" is a topic label. "Your phone is the reason you can't focus" is a promise. One of those gets tapped. The cover exists entirely to earn the first swipe — nothing else.

Rule 2: Slide 2 confirms the hook or they leave

The second slide is where you either validate the decision to keep reading or lose the person permanently. It doesn't explain the topic — it opens the problem.

Set up the stakes. Make the reader think: this is worth my time. If slide 2 restates the cover or gives background, the reader leaves. If it immediately makes the problem feel real and urgent, they stay.

Rule 3: Every slide works as a screenshot

The test: pull any single slide out of the carousel and show it to someone with no context. Does it still land?

Each slide should carry one complete idea. No slide should require the reader to remember what the previous slide said. This is what makes content saveable — any slide can be the entry point.

Rule 4: One slide must be worth saving

Every carousel needs a save slide. A framework. A fill-in-the-blank template. A checklist. A line sharp enough to quote.

If someone has to save the whole carousel to access the value, you've buried it. The save slide should be identifiable — the reader should know within two seconds whether this is the thing they wanted to keep.

Rule 5: Stop writing from scratch

AI handles the structure. You handle the taste.

The fastest carousel workflow: give an AI model your topic, your angle, and your brand context. It returns a slide-by-slide structure with the job of each slide named. You then edit from that output — adjusting voice, sharpening lines, injecting specifics. What takes an hour now takes ten minutes.

Rule 6: The caption is a second post — the CTA is the engine

The caption isn't a summary of what the carousel said. It restates the core idea and adds one piece of context the slides didn't have. Treat it as a standalone piece of writing that works even if someone never taps the carousel.

The CTA isn't "follow me for more." It's a comment trigger: "Comment [KEYWORD] and I'll send it to you." This drives replies, which drives reach, which is the actual engine.

Prompt 1 — The Carousel Architect

This prompt generates a complete slide-by-slide carousel structure. Paste your topic, angle, and brand context. Edit the output — don't post it raw.

promptCopy
You are my carousel architect. I'll give you a topic, an angle, and my brand context. Return a complete slide-by-slide structure for an Instagram carousel.

Follow these 6 rules without exception:
1. The cover stops the scroll. Max 8 words. A promise or a number, never a topic label.
2. Slide 2 confirms the hook. Set up the problem and the stakes so the reader thinks this is worth my time.
3. Every slide works alone. One idea per slide. A slide screenshotted with no context still lands.
4. One slide is save-worthy. A framework, a prompt, a fill-in-the-blank, or a line sharp enough to quote.
5. The structure carries the narrative. Cover, confirm, build, climax, CTA. Energy rises, then resolves.
6. The CTA slide names one action and one reward.

My input:
- Topic: [TOPIC]
- Angle: [THE SPECIFIC TAKE I HAVE THAT MOST PEOPLE DON'T]
- Brand context: [WHO IT'S FOR, MY VOICE, MY POSITIONING]

Your output:
- 7 to 10 slides total
- For each slide: the number, its job (cover / confirm / build / climax / CTA), and the exact copy
- Flag which slide is the save-worthy one
- One idea per slide, no clutter
- Match my voice. No hustle language, no platitudes.

Give me the structure. I'll edit from there.

Prompt 2 — The Cover Line Generator

Run this after Prompt 1. It takes your carousel structure and returns 10 cover line options ranked from safest to boldest. Pick one. Don't overthink it.

promptCopy
Here's a carousel structure. Give me 10 cover line options. The cover has one job: stop the scroll.

- Max 8 words
- A promise or a number, not a topic label
- Short enough to be one bold line on a slide
- It either gets saved or gets argued with

Carousel structure: [PASTE THE STRUCTURE FROM PROMPT 1]
Audience: [WHO THIS IS FOR]

Return 10 options, sorted from safest to boldest. No explanations.

Prompt 3 — The Caption Writer

Run this last. The caption is the second post — not a recap of the carousel. It opens with a sharp claim, restates the core idea with new context, and ends with a comment CTA.

promptCopy
Turn this carousel into an Instagram caption in my voice. The caption is the second post, not a recap.

It must:
- Open with a sharp claim or a felt frustration
- Restate the core idea and add one piece of context the carousel didn't have
- Stay under 150 words
- End with: Comment [KEYWORD] and I'll send it to you.

Carousel: [PASTE THE SLIDES]
Keyword: [YOUR KEYWORD, IN CAPS]
Voice: [DIRECT, DECLARATIVE, NO PLATITUDES, NO HUSTLE LANGUAGE]

End with my sign-off line.

Pre-flight checklist before posting

• Cover: 8 words or fewer. A promise or a number, not a topic label.
• Slide 2: Opens a problem or stakes, doesn't restate the cover.
• Screenshot test: Pull any inner slide. Does it still make sense with no context?
• Save slide: Is there one slide worth bookmarking on its own?
• Visual system: Do all slides use your established brand colours, fonts, and layout?
• Caption and CTA: Caption adds context the carousel didn't have. CTA is a comment trigger with a keyword.

Frequently asked questions

How many slides should a carousel have? Seven to ten. Fewer than seven rarely builds enough narrative arc to earn a save. More than ten starts to feel like homework. The Architect prompt targets that range by default. If you're consistently writing 12-slide carousels, audit your structure — you're likely duplicating ideas across slides or including context that belongs in the caption, not the carousel.

What makes a good comment trigger keyword? Short, specific, and directly tied to what you're sending. "AI" is too broad. "SYSTEM" is better. "TEMPLATE" works if the thing you're sending is a template. The keyword should be something the reader would type without hesitation — which means it should match exactly what they feel they're getting in return. Broad keywords get ignored; specific ones get comments.

Should I edit the AI output or post it directly? Always edit. The Architect prompt returns a structure, not a finished carousel. What AI gives you is the architecture — which slides exist, what job each one does, a rough version of the copy. What you add is voice: the specific words that sound like you, the examples only you would use, the takes only your experience produces. Post the raw output and it reads like everyone else's content. Edit it and it reads like yours.