Carousels live or die on slide 1 (the hook) and the last slide (the save). These 12 prompts walk you slide-by-slide through structure, pacing, and the save-driving payoff most carousels miss. Each is tagged [IG], [LinkedIn], or [Both] so you write to the right algorithm the first time.
TL;DR
- Carousels are the highest-engagement organic format on both platforms right now: on Instagram they out-reach single images by roughly 3x and double the saves of Reels; LinkedIn document posts hit ~6.6% engagement, the top format on that feed (as of June 2026).
- Instagram weights saves and shares; LinkedIn weights dwell time and comments. Write slide 1 and the last slide to those two different goals.
- Specs that matter: Instagram allows up to 20 slides (raised from 10), 4:5 portrait at 1080×1350. LinkedIn document posts accept PDF/PPTX/DOCX up to 100 MB and 300 pages.
- Feed any model the prompts below. GPT-5.5, Claude Sonnet 4.6, and Gemini 3.1 Pro all handle slide-by-slide structure well; paste your draft back in for the retention check (prompt 12).
IG carousel vs LinkedIn document post: what actually differs
The two formats look identical from a distance, a stack of slides you swipe through, but the algorithm each lives inside is not the same, and the slide that wins on one platform often loses on the other. Below, the same topic (“5 indie-maker mistakes”) rebuilt for each.
Instagram carousel
- Aspect ratio: 1080×1350 (4:5 portrait fills the most screen without a zoom)
- Slide count: 8-10 sweet spot; hard cap is 20 slides as of June 2026
- Algorithm: SAVES and SHARES over likes; Instagram also re-serves high-swipe carousels 24-48 hours later, showing slides a user skipped the first time
- Slide 1: pattern interrupt, bold claim, oversized text, no logo clutter
- Middle slides: visual-led, short text overlay, generous whitespace
- Last slide: CTA (save, share, follow) + handle
Sample IG-coded template for “5 indie-maker mistakes”:
Slide 1 (cover, pattern interrupt):
"I lost $40k to these 5 mistakes."
Huge text, off-center, no logo.
Slide 2 (the promise):
"All 5 - and the fix - by slide 7. Save it."
Slides 3-7 (one mistake per slide):
Mistake headline (4-5 words, oversized)
One-line consequence
One-line fix
No paragraphs. Whitespace over info density.
Slide 8 (save CTA):
"Save this. You'll do at least 2 of these in year 1."
Handle bottom-right. No headshot.LinkedIn document post
- Aspect ratio: 1080×1350 (4:5) is best for mobile; 1:1 and 16:9 also render; PDF/PPTX/DOCX up to 100 MB, 300 pages
- Slide count: 8-12 (LinkedIn readers absorb more per slide); aim for under ~30 words per slide
- Algorithm: DWELL TIME first, then COMMENTS; a click-through reads as high-value and widens reach
- Slide 1: professional headline (no clickbait), clear thesis
- Middle slides: data + insight, more text per slide, restrained design (less playful than IG)
- Last slide: a question to drive comments + handle + headshot
Sample LinkedIn-coded template for “5 indie-maker mistakes”, same topic, different execution:
Slide 1 (cover, thesis):
"5 mistakes most indie makers make in year 1 - and what the data shows."
Clean type, small headshot bottom-left.
Slide 2 (context):
Who I am, sample size (e.g. "Reviewed 200 indie SaaS launches in 2024-25").
Slides 3-12 (one mistake per 2 slides):
Slide A: Mistake + the data point that proves it
Slide B: The fix + a 1-line example
Paragraphs OK. Cite where the data came from.
Slide 12 (comment CTA):
"Which of these did you do? Which would you add?"
Handle + headshot + small logo.The shorthand: IG slide 1 has to stop the thumb; LinkedIn slide 1 has to earn the dwell. IG ends with “save this”; LinkedIn ends with “what’s your take”.
| Spec | Instagram carousel | LinkedIn document post |
|---|---|---|
| Top-weighted signal | Saves and shares | Dwell time, then comments |
| Slide sweet spot | 8-10 | 8-12 |
| Hard cap | 20 slides | 300 pages / 100 MB |
| Best aspect ratio | 4:5 (1080×1350) | 4:5 (1080×1350) |
| Slide-1 job | Stop the thumb | Earn the dwell |
| Last-slide job | Drive a save | Drive a comment |
| Tone | Playful, big type | Professional, data-led |
When you use the 12 prompts below, each is tagged with its best-fit platform: [IG], [LinkedIn], or [Both]. Replace each bracketed placeholder (for example [topic], [platform], [N]) with your own input before running.
Best for
- Instagram carousels and LinkedIn document posts
- Educational and personal-brand storytelling
- Repurposing one long-form post into 8-12 slides
1. 8-slide educational carousel [Both]
Design an 8-slide educational carousel on "[topic]". Output per slide: headline, 1-line body, visual cue. Slide 1: cover hook. Slide 2: "this is why it matters". Slides 3-6: 4 specific points. Slide 7: recap. Slide 8: CTA / save.2. Cover-slide A/B variants [Both]
Generate 6 cover slide variants for a carousel on "[topic]". Each: a hook headline of 7 words or fewer, a sub-line, a visual cue. Vary the angle: contrarian, numbered, question, "stop doing X". Mark the strongest for [platform].3. Story-arc carousel (10 slides) [LinkedIn]
Design a 10-slide story carousel. Story: [paste]. Output: slide 1 hook, slides 2-3 setup, slides 4-6 conflict, slides 7-8 turning point, slide 9 lesson, slide 10 CTA. Each slide 25 words of body or fewer.4. Listicle carousel [Both]
Design a "[N] [things] for [audience]" carousel. Output: cover, intro slide, N item slides each with the same structure (headline / 1-line body / 1 mini-example), outro slide with save-CTA.5. Before / after carousel [IG]
Design a 7-slide "before / after" carousel for [topic]. Output: cover, before state (2 slides), the turning point (1 slide), after state (2 slides), the principle behind it (1 slide).6. Mistake-fix carousel [Both]
Design a 9-slide carousel: "[N] mistakes in [field] (and how to fix)". Slides: cover, intro context, then for each mistake one slide describing it plus one slide with the fix. End with a saveable summary slide.7. Tutorial / how-to carousel [Both]
Design an 8-slide how-to carousel for [task]. Slides: cover, prerequisites, 5 step slides, outro with "save this for later" plus a sub-CTA.8. Mental-model / framework carousel [LinkedIn]
Design a 7-slide carousel introducing a framework for [decision]. Slides: cover, the problem, the framework name plus diagram description, 3 application slides, save-CTA.9. Repurpose long-form to carousel [LinkedIn]
Below is a 1,500-word article. Extract a 10-slide carousel that preserves the strongest argument. Mark which paragraph in the source becomes each slide. Cut the rest ruthlessly.
[paste article]10. Stat-led data carousel [LinkedIn]
Design a 6-slide carousel built around a striking statistic about [topic]. Slides: cover with the number, what the number means, 2 examples of it, what to do about it, save-CTA. Source the statistic explicitly.11. Save-driving outro slide [IG]
My carousel has a weak final slide. Below: the topic and slide 1. Write 5 strong final-slide variants that drive saves: summary, "screenshot this", framework recap, mini-checklist, decision tree.
[paste topic + slide 1]12. Slide-by-slide retention check [Both]
Below is my full carousel. Walk through each slide and predict where readers will swipe away vs keep going. Suggest the rewrite that holds attention through the last slide.
[paste slides]Common mistakes
- A cover slide that does not promise a payoff (the single biggest reach killer on both platforms)
- Burying the lead in slide 4 instead of teasing it on slide 2
- A last slide that says “follow me for more” with no save or comment value
- Inconsistent visual style or mixed aspect ratios across slides (Instagram penalizes mixed ratios)
- Too many slides: past ~10 on Instagram or ~12 on LinkedIn, completion rate drops sharply
FAQ
How many slides should a carousel have? Treat 8-10 as the default on Instagram and 8-12 on LinkedIn. Instagram now allows up to 20 slides, but completion rate falls off after about 10 unless every slide earns its place. LinkedIn rewards dwell time, so a slightly longer, denser deck can work if each slide stays under roughly 30 words.
Which AI model writes the best carousel copy? All three frontier models handle the slide-by-slide structure in these prompts. Claude Sonnet 4.6 tends to keep slide bodies tight and on-format; GPT-5.5 is strong at punchy cover hooks; Gemini 3.1 Pro is fine for first drafts. Whatever you use, run prompt 12 (the retention check) on the output before you design it.
Does the AI design the slides too? No. These prompts produce the copy and structure (headline, body, visual cue per slide). Move the output into a design tool such as Canva or Figma to lay out the slides. Keep one aspect ratio and one type system across the whole deck.
Why does Instagram weight saves but LinkedIn weights comments? Instagram reads a save as “I want to come back to this”, a strong long-term value signal, and re-serves high-swipe carousels 24-48 hours later. LinkedIn optimizes for dwell time and conversation, so a comment (and the back-and-forth it triggers) widens reach more than a silent save. Write your last slide to whichever signal the platform rewards.
Can I reuse the same carousel on both platforms?
Reuse the topic and core points, not the execution. The same “5 mistakes” deck needs a thumb-stopping cover and a save CTA on Instagram, and a thesis cover plus a comment-driving question on LinkedIn. Prompts 1, 2, 4, 6, 7, and 12 are tagged [Both] for exactly this reason.