Social Media Carousel Prompt is a Claude AI skill — diagnoses format issues before writing, building platform-calibrated slide architecture inside Claude.
Here's the frustrating part: most carousel posts that underperform aren't badly written. The ideas are coherent. The copy is clear. Someone spent real time on it. And it still gets ignored — 80 impressions, no saves, three people who opened it and swiped to slide two before dropping off. Meanwhile, an account with a fifth of your following posts something that structurally looks similar and it runs to 400 saves and a few hundred new followers over the next week.
The gap is almost never about the quality of the writing. It's about slide architecture — specifically, whether the deck is built on a format that matches how people navigate carousel content on that platform right now. Carousel format patterns shift. What earned saves on LinkedIn 18 months ago reads as stale today. What works on Instagram educational content is structurally different from what works on a thought-leadership carousel targeting the same demographic. And what vanilla AI generates when you ask for a carousel is built from training data that doesn't reflect any of that.
This is a diagnostic post. Each section below names one structural failure pattern, shows you what it looks like in practice, and explains what actually fixes it.
The Four Patterns Killing Your Carousel Before Slide Three
Pattern 1: The Overused Hook Format
Hook formats have a lifespan. A carousel that opens with "X things I wish I knew about Y" worked at scale on LinkedIn until it got replicated so widely that the format itself became the signal to scroll. The same happened with the "hot take + numbered list" structure on Instagram educational content. When enough accounts use the same hook architecture, the format stops reading as content and starts reading as template — and the algorithm treats it accordingly, because dwell time drops immediately.
Open your last five carousels and look only at slide one. If the hook format appears across multiple other accounts you follow — as a recognisable pattern, not just a similar topic — it's likely stale. The question isn't whether your hook is well-written. It's whether the format it uses still reads as fresh on your platform.
What replaces an overused hook format isn't another format — it's the format that's currently outperforming on your specific platform and content category. That changes faster than most content calendars account for, which is why live research before writing matters more than any fixed template.
| Stale hook pattern | What it signals to the reader |
|---|---|
| "X things I wish I knew about Y" | This deck is a list. I already know what's in it. |
| "Most people do X wrong. Here's how to fix it." | Generic authority claim. Overused on every niche. |
| "Stop doing X. Start doing Y instead." | Contrarian-bait format. Peaked 2022–2023. |
| "I spent [time] learning X so you don't have to." | Research-proxy format. Now reads as AI-generated. |
Pattern 2: Information Density Mismatch
Every platform has an information density norm for carousel content — how much text and how much visual breathing room each slide should have. LinkedIn's professional audience tolerates higher text density per slide than Instagram's visual-native users. B2B educational carousels on Instagram can carry more copy per slide than personal brand or lifestyle content. A slide deck calibrated to the wrong density norm creates friction: either the audience feels the deck is too thin to be worth the swipe, or each slide requires more cognitive effort than the platform's average attention span supports.
Using the same slide template across LinkedIn and Instagram because the audience demographic overlaps. Platform norms for information density are set by how people physically interact with the feed on each app — not by who the audience is. A LinkedIn carousel that performs well will almost always be too dense for the same content on Instagram.
The fix isn't to write less (or more). It's to calibrate density to what's currently performing on the specific platform for your content category. A skill that researches current format norms before building slide architecture handles this automatically — the density calibration is baked in before a single word is written.
Pattern 3: The Momentum Drop at Slide Four
Most carousel decks that lose readers lose them at slide four. Not because slide four is bad — but because the structural arc of the deck gives the reader permission to stop. Slides one through three delivered the promise of the hook. Slide four is where the deck either builds genuine new tension ("there's something you haven't seen yet") or starts delivering the list items the reader already expected. When the reader can predict what slides five through ten contain from slide four, there's no structural reason to keep swiping.
Read your last carousel without looking at the platform analytics. At slide four, ask: does the reader know what's coming next? If yes — if the arc is that predictable — you've built a list, not a story. Lists get read once and not saved. Stories get saved because the reader wants to return to them.
Strong carousel architecture places a structural pivot between slide three and slide five — a reframe, a revelation, a complication that changes the stakes of what the earlier slides established. This is the hardest element of carousel writing to produce with a generic AI approach, because it requires understanding both the reader's current expectation and the platform's current content grammar. A static approach to slide structure doesn't know either.
Pattern 4: The Save-Trigger That Doesn't Trigger
The final slide of a carousel is the save mechanic. It's the answer to the question every reader is implicitly asking: "Why would I keep this?" A final slide that says "Which of these tips resonated most? Comment below!" is optimised for comments, not saves — and on most platforms those aren't equivalent outcomes. A final slide that delivers a concise, high-density summary of the framework the deck just taught is optimised for saves, because the reader wants to return to it when they're applying the information.
| Final slide approach | What it's actually optimised for |
|---|---|
| "Comment which tip was your favourite" | Engagement bait — optimised for comments, not saves or shares |
| "Follow for more content like this" | Follower acquisition — not a save trigger |
| "Save this for later" with no reason to | Asks for the save without earning it — rarely works |
| One-slide reference summary of the deck's core framework | Optimised for saves — gives the reader a reason to return |
| A single actionable next step the reader can take today | Optimised for saves and shares — specific utility drives both |
The save-trigger is earned by the content of the deck, not the copy of the final slide. A deck that teaches something the reader wants to reference again doesn't need to ask for the save. The final slide should be built as the artefact the reader keeps — a dense, referenceable summary of what they just learned, not a request for engagement.
That gap is exactly what the Social Media Carousel Prompt skill for Claude was built to close.
What Platform-Aware Structure Changes
The four patterns above share a root cause: they're all the result of building a carousel deck without current platform intelligence. A generic AI approach to carousel writing applies the same slide structure regardless of platform, audience, or what's currently performing in the feed. It can't do otherwise — the training data doesn't contain this week's format norms.
The NovaKit Social Media Carousel Prompt skill addresses this at the structural level by researching current format patterns on your specified platform before building the slide architecture. The hook format, information density, structural arc, and final-slide mechanic are all calibrated to what's working right now — not to a template derived from content that was performing 18 months ago.
The three-question calibration (platform, niche, core idea) means the output is also built for your specific content category, not just your platform. A B2B software LinkedIn carousel and a personal brand thought-leadership LinkedIn carousel call for different structural logic. The skill distinguishes between them because it asks before it builds.
The creator who keeps asking "why isn't my carousel working?" is usually asking the wrong question. The deck that earns saves isn't the one with better writing — it's the one built to the right architecture for the right platform at the right moment. Format patterns are infrastructure, not decoration. When the infrastructure is wrong, better copy can't fix it. It just makes the wrong structure more polished.
If you're doing this manually, carousel prompts designed for swipe-through completion is the faster path.
Put this to work: the Social Media Carousel Prompt skill for Claude turns everything above into one guided workflow you run in a normal Claude chat. Not ready to buy? Start with a free Claude skill and see how it works first.