ChatGPT prompts for carousel outlines, hooks and captions
Slide outlines, hook-first covers and caption packs you paste into a PDF. Use facts you will stand behind. Distinct from the Claude carousel page and from LinkedIn profile prompts.
In short: This page contains 30 copy-paste ready prompts, organized into 6 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Drafted with AI writing assistance and checked against our editorial policy.
Hook First Slides
5 promptsCover line from a messy note
1/30✨ What it does
Offers five 6 to 10 word covers from your notes, then picks the one that would stop a tired scroller. Covers that need a missing number stay tagged.
Write five LinkedIn carousel cover options of 6 to 10 words from notes I will stand behind. Topic: [TOPIC]. Audience: [AUDIENCE]. Raw notes: [PASTE NOTES]. Output a numbered list, then pick the cover that would stop a tired scroller in 40 words. If a number, client or result is not in the notes, write [VERIFY] on that cover and do not invent a statistic. Return covers only, no design notes, no HTML.
Pro tip: Paste the messy note, not a polished post. Polish hides the hook.
Question cover the feed already asks
2/30✨ What it does
Turns a question you actually hear into three short covers and a slide-two tease. Missing rates stay tagged instead of becoming fake proof.
Turn this recurring question into a carousel cover and a slide-two tease. Question I actually hear: [QUESTION]. Who asks it: [ASKER]. What I truly answer with: [MY ANSWER IN BRIEF]. Write three cover lines as questions under 12 words, then one slide-two sentence that promises the method without a number I did not give. If I did not supply a rate or a count, put [VERIFY] rather than inventing a figure. No HTML. Caption later, not now.
Pro tip: Use the asker's words. Cleaned-up jargon does not stop a thumb.
Before and after cover, facts only
3/30✨ What it does
Builds three before-and-after cover pairs from states you can defend, then recommends one. Missing metrics stay blank and tagged.
Draft a two-line LinkedIn cover that contrasts before and after using only my facts. Before state: [BEFORE]. After state: [AFTER]. Time span I can defend: [TIME SPAN]. Output three cover pairs (line A / line B), each line under 8 words, then recommend one pair in 30 words. If the after state needs a metric I did not give, write [VERIFY] and leave the number blank. Do not invent a client name or a percent. Slides only.
Pro tip: If you cannot name the time span, leave it out of the cover.
Myth on the cover, cost on slide two
4/30✨ What it does
Opens on a myth in the reader's words and puts the cost you have seen on slide two. Implied figures you did not type stay tagged.
Write a carousel open that names a myth I actually hear, then the cost I have seen. Myth: [MYTH IN THEIR WORDS]. Cost I can stand behind: [COST IN MY WORDS]. Audience: [AUDIENCE]. Return: 1 cover line under 10 words, 1 slide-two body of 20 to 28 words, and 1 alt cover. If the cost implies a dollar amount, a percent or a headcount I did not type, mark it [VERIFY] and skip inventing the figure. No scare claims I did not write.
Pro tip: Quote the myth the way a client said it, not the way a textbook would.
Cover that names the skipped step
5/30✨ What it does
Gives four covers that name the skipped step rather than the whole process, plus a short reason. Covers that would need a failure rate stay tagged.
Build a hook-first cover around the step people skip in [PROCESS]. Skipped step: [SKIPPED STEP]. What breaks: [BREAK IN MY WORDS]. Who it hurts: [WHO]. Give four covers under 9 words that name the skip, not the whole process, then one 25-word reason for your favorite. If I did not give a failure rate, do not invent one; write [VERIFY] beside any cover that would need it. Output covers and the reason, no Canva notes.
Pro tip: Name the skip. Process improvement as a cover is invisible in the feed.
Slide Outlines
5 promptsEight-slide how-to skeleton
6/30✨ What it does
Returns an 8-slide how-to skeleton with a cover, one listed step per slide and a closer. Missing tools or times are tagged, not invented.
Outline an 8-slide LinkedIn how-to carousel I will design myself. Process: [PROCESS]. Ordered steps I actually run: [STEPS]. Audience: [AUDIENCE]. For each slide return Headline (max 8 words) and Body (20 to 30 words). Slide 1 is the cover promise. Slides 2 to 7 are one step each from my list. Slide 8 is a takeaway plus [CTA]. Do not add a step I did not list. If a tool, time or count is missing, write [VERIFY] in that body rather than inventing it. No HTML.
Pro tip: If slide 2 restates the cover, rewrite it before you open the design file.
Seven-slide beginner walkthrough
7/30✨ What it does
Walks a beginner through seven slides to one named today-action using tools they already have. Undefined terms stay tagged.
Outline a 7-slide beginner carousel for [SKILL] that ends on one action they can take today. Starting point: [STARTING POINT]. Tools they already have: [TOOLS]. Today action: [TODAY ACTION]. Voice: [PATIENT OR DIRECT]. Each content slide: headline under 8 words, body 18 to 25 words. Define any term from [TERMS] in plain words on the slide it first appears. Do not invent a tool feature I did not name. If a term lacks a definition in my input, tag it [VERIFY]. Caption not required in this pass.
Pro tip: Assume they have never done this. Advanced asides belong in a later deck.
Checklist carousel, binary lines
8/30✨ What it does
Turns your checks into binary yes-or-no slides plus a cover on what the list prevents. Costs that would need a number stay tagged.
Turn my checks into a saveable LinkedIn checklist carousel. Project: [PROJECT]. Checks I run: [CHECKS]. The check people skip: [SKIPPED CHECK]. Audience: [AUDIENCE]. Write 8 to 10 slides: cover on what this list prevents, then one binary check per slide (headline as a yes or no), then a closer with [CTA]. Body 15 to 22 words on the cost of skipping. Do not invent a check. If a cost would need a number I did not give, write [VERIFY] instead of inventing a stat.
Pro tip: Each check must be binary. Be thoughtful is not a line people save.
Click-path slides from my notes
9/30✨ What it does
Maps each click to a slide whose headline is the button name, plus a missed-click callout. Fuzzy labels stay tagged instead of guessed.
Outline a click-path carousel for [TOOL AND TASK]. Click path in my words: [CLICK PATH]. Done state: [DONE STATE]. Missed click: [MISSED CLICK]. Audience: [AUDIENCE]. One slide per click. Headline equals the button or menu name. Body 20 to 28 words on what to look for. Include a cover and a missed-click callout. Do not invent a menu I did not name. If a label is fuzzy, write [VERIFY] on that headline rather than guessing the UI. No screenshots, copy only.
Pro tip: Headline the control. Think strategically as a title cannot be followed.
Repair-the-process outline
10/30✨ What it does
Names the break, the old path, your repair steps and what still breaks. Missing costs stay tagged and off-limits claims stay out.
Outline an 8-slide repair carousel. Broken process in their words: [BROKEN PROCESS]. Cost I can defend: [COST]. Repair steps I have run: [REPAIR STEPS]. Off limits: [OFF LIMITS]. Audience: [AUDIENCE]. Cover names the break. Then old path in one slide, one repair step per slide, what still breaks, closer ask. Headline under 8 words, body 20 to 30 words. Honor off limits. If a cost or timeline is not in my input, mark [VERIFY] and do not invent a cleaner ending than I gave.
Pro tip: Name the break the way a teammate would slack it, not as inefficiency.
Caption Packs
5 promptsCaption with a messy aside
11/30✨ What it does
Writes a 250 to 400 word caption that teases the swipe, adds your messy aside and ends on a question. Missing metrics stay tagged.
Write a LinkedIn caption of 250 to 400 words for this carousel. Slide headlines I will post: [SLIDE HEADLINES]. Messy aside that did not fit a slide: [ASIDE]. Audience: [AUDIENCE]. Tease the swipe in the first two lines, then the aside, then one comment question. Mark a [SEE MORE] cut after line two. Use only facts in the headlines and aside. If you want a number I did not give, write [VERIFY] rather than inventing a metric. No hashtag stuffing. No invented client quote.
Pro tip: Put the aside after See more. Repeating every slide in the caption wastes the swipe.
Short caption plus alt covers
12/30✨ What it does
Gives an 80 to 120 word caption and two short first-comment options from proof you can show. Thin proof stays tagged, not padded.
I need a short caption (80 to 120 words) and two alternate first comments for the same carousel. Topic: [TOPIC]. Promise on the cover: [COVER LINE]. One proof I can show: [PROOF]. CTA: [CTA]. Keep the caption human, one idea. If proof lacks a figure, tag [VERIFY] and do not invent a percent to make it punchy. Do not invent a colleague who agrees. Output caption, then Comment A, Comment B, each under 40 words.
Pro tip: Write the first comment yourself later if the model still sounds like a billboard.
Caption from a story I lived
13/30✨ What it does
Turns a lived note into a 200 to 350 word caption that leaves the slides to do the teaching. Missing endings and results stay tagged.
Turn this lived note into a caption that does not retell every slide. Note: [PASTE STORY]. Slides will cover: [SLIDE PROMISE]. Audience: [AUDIENCE]. 200 to 350 words. Keep people as a client or a teammate, no real names unless I typed them. Do not invent an ending I did not write. If a result is missing, put [VERIFY] where the result would go. End with a question that invites their version, not a humblebrag. No HTML.
Pro tip: Strip names before you paste if the person has not agreed to be in a post.
Three caption lengths, same facts
14/30✨ What it does
Produces 50, 150 and 300 word captions from the same facts, adding texture only in the long one. New claims that would need a stat stay tagged.
Write three captions from the same facts: 50 words, 150 words, and 300 words. Facts: [FACTS]. Cover line: [COVER LINE]. CTA: [CTA]. Keep the facts identical across lengths; the long one may add texture, not new claims. If a length would need a stat I did not supply, write [VERIFY] in that version and do not invent it. Label each caption. No extra case studies. No fake awards.
Pro tip: Post the 150 unless the story truly needs the 300. Long captions hide the PDF.
Caption question bank
15/30✨ What it does
Lists eight specific comment questions tied to your slide ideas, with no invented poll results. Questions that assumed a number get tagged and rewritten.
Give me eight comment-starting questions I can rotate under carousels on [THEME]. Each question under 18 words, specific to [AUDIENCE], not 'thoughts?'. Tie each question to a slide idea from [SLIDE IDEAS]. Do not invent a poll result or a percent of readers. If a question assumes a number I did not give, tag it [VERIFY] and rewrite it without the number. Output a numbered list only, then one line on which question fits a how-to versus a story deck.
Pro tip: Ask which step they skip. Yes or no questions die in the comments.
Like these prompts? There are full tutorials behind them.
Learn the workflows, not just the prompts. Easy-to-follow tutorials inside AI Academy, and the library keeps growing.
Story Carousels
5 promptsEight-beat lived story deck
16/30✨ What it does
Maps a lived situation onto eight beats from belief through lesson. Missing outcome numbers stay tagged and off-limits claims stay out.
Outline an 8-slide story carousel from a situation I lived. Situation: [SITUATION]. What I believed: [BELIEF]. What broke: [BREAK]. What I did: [ACTION]. What I will not claim: [OFF LIMITS]. Audience: [AUDIENCE]. Beats: cover, belief, break, action, messy middle, outcome only if I stated one, lesson, CTA [CTA]. Headline under 8 words, body 20 to 30 words. Do not invent a neater ending. If an outcome number is missing, write [VERIFY] on that slide and keep the lesson qualitative.
Pro tip: If the outcome is still in progress, say so on the slide. Fake endings get called out.
Client story with names stripped
17/30✨ What it does
Outlines seven slides from a client note with identifiers stripped and identifying metrics flagged. Results they did not give stay out.
Turn this client note into a carousel outline that cannot identify them. Note: [PASTE NOTE]. Role I will use instead of a name: [ROLE]. Lesson: [LESSON]. Audience: [AUDIENCE]. 7 slides: cover, context without identifiers, the ask, the constraint, the move, the lesson, CTA. Body 20 to 28 words. Remove company, city and unusual stack details. Do not invent a result they did not give. If a metric is in the note but looks identifying, replace it with [VERIFY] and ask me whether to keep it.
Pro tip: Get a yes before you post even an anonymized story if the constraint is unique.
Failure post as a carousel
18/30✨ What it does
Builds an honest failure deck from the miss, the assumption, the cost and the change. Unsupplied dollar figures stay tagged, and secret wins stay out.
Outline a failure carousel I am willing to own. What failed: [FAILURE]. What I had assumed: [ASSUMPTION]. Cost I can state: [COST]. What I changed: [CHANGE]. Audience: [AUDIENCE]. 8 slides, honest, not a humblebrag that secretly wins. Do not invent a later win I did not write. If the cost is not a number I supplied, mark [VERIFY] rather than inventing a dollar figure. Headline under 8 words. Closer: one question about their similar miss, plus [CTA].
Pro tip: If you later won, put that in a different deck. Mixing it in reads as a setup.
Week-in-the-work story
19/30✨ What it does
Turns week notes into six behind-the-work slides with one copyable move. Missing counts stay tagged and a dramatic Friday is not invented.
Turn my week notes into a 6-slide behind-the-work carousel. Notes: [WEEK NOTES]. Role: [ROLE]. What I want them to copy: [COPYABLE MOVE]. Audience: [AUDIENCE]. Slides: cover, Monday constraint, the ugly middle, the move, what I would repeat, CTA. Use only events in the notes. Do not invent a dramatic Friday. If a time spent or a count is missing, write [VERIFY] in that body. Keep it peer-to-peer, not a productivity lecture.
Pro tip: Pick one copyable move. A whole week of lessons will not fit six slides.
Two-timeline then and now
20/30✨ What it does
Alternates then and now slides from facts you can defend, with unmatched metrics tagged. Bridge years you did not live stay out.
Outline a then-and-now carousel. Then, facts I can defend: [THEN FACTS]. Now, facts I can defend: [NOW FACTS]. What stayed the same: [SAME]. Audience: [AUDIENCE]. 8 slides alternating then and now, plus a cover and a closer. Do not invent a bridge year. If a comparison needs a metric I gave on only one side, tag the other side [VERIFY] instead of inventing a match. Headlines under 8 words. No fake origin myth.
Pro tip: If now is smaller than then, say so. Inflated now-slides are easy to spot.
List Carousels
5 promptsSeven mistakes I still see
21/30✨ What it does
Lists seven mistakes you actually see with a short instead-do body from your fix notes. Unlisted mistakes and fake success rates stay out.
Outline a 9-slide list carousel: cover, seven mistakes, closer. Theme: [THEME]. Mistakes I actually see: [MISTAKES]. Who sees them: [AUDIENCE]. Each mistake slide: headline is the mistake in their words, body 18 to 25 words on what to do instead from [FIX NOTES]. Do not invent a mistake I did not list. If a fix would need a stat, write [VERIFY] rather than inventing a success rate. CTA: [CTA]. No dunking on beginners.
Pro tip: Seven is enough. Ten-mistake decks become a rant people mute.
Five questions I ask before I start
22/30✨ What it does
Puts each question you actually ask on its own slide with a qualitative why. A sixth question is not added, and missing numbers stay tagged.
Write a carousel of questions I actually ask before [WORK TYPE]. Questions: [QUESTIONS]. Why each matters, in my words: [WHY NOTES]. Audience: [AUDIENCE]. Cover plus one question per slide plus closer [CTA]. Headline is the question, under 12 words. Body 20 to 30 words. Do not invent a sixth question. If a why would need a number I did not give, mark [VERIFY] and keep the why qualitative. Peer tone, not a gatekeep quiz.
Pro tip: Headline the question. A label like discovery does not get saved.
Tools I actually open, not a stack flex
23/30✨ What it does
Gives one slide per tool you truly use, headlined by the job and honest about a limit. Rankings and user-counts you did not give stay out.
Outline a tools carousel from tools I truly use. Tools: [TOOLS]. Job each does for me: [JOB PER TOOL]. Audience: [AUDIENCE]. Cover that promises the job, not a stack. One tool per slide, headline is the job, body names the tool and one limit I accept. Closer [CTA]. Do not invent a tool. Do not invent a user-count or a ranking. If I skipped a limit, write [VERIFY] on that slide. No affiliate tone.
Pro tip: Headline the job. Tool names as headlines look like ads.
Red flags in [SITUATION]
24/30✨ What it does
Lists hard-way flags with your response and one not-a-flag slide. Horror stories and percents you did not supply stay out.
Write a red-flag carousel for [SITUATION]. Flags I have learned the hard way: [FLAGS]. What I do when I see one: [RESPONSE]. Audience: [AUDIENCE]. Cover, one flag per slide, a 'this is not a flag' slide from [NOT A FLAG], closer. Body 18 to 25 words. Do not invent a lawsuit, a percent or a horror story. If a flag needs a detail I did not give, tag [VERIFY]. Firm and kind, not a pile-on.
Pro tip: Include a not-a-flag slide. All-red decks train people to panic at normal work.
Resources I send, with a caveat each
25/30✨ What it does
Puts each resource you actually send on a slide with its caveat. Missing caveats stay tagged, and star ratings are not invented.
Outline a resources carousel. Resources I actually send: [RESOURCES]. Caveat for each: [CAVEATS]. Who I send them to: [AUDIENCE]. Cover, one resource per slide with its caveat, closer [CTA]. Headline under 8 words. Body 20 to 28 words. Do not invent a resource or a star rating. If a caveat is missing, write [VERIFY] on that slide rather than inventing a warning. Not a dump of 30 links.
Pro tip: One caveat per resource. Unqualified love-lists get ignored by serious readers.
Go from copy-pasting to actually mastering AI.
AI Academy: hands-on tutorials on ChatGPT, Claude, Midjourney, and many other tools. New tutorials are added regularly.
Framework Decks
5 promptsNamed three-part method
26/30✨ What it does
Turns your method into cover, definition, one slide per part, a situation and a do-not-use. Extra parts and unsourced metrics stay out.
Turn my working method into a named-framework carousel. Name I will stand behind: [FRAMEWORK NAME]. Parts in my words: [PARTS]. One-line definition: [DEFINITION]. A real situation it explained: [SITUATION]. When not to use it: [DO NOT USE WHEN]. Audience: [AUDIENCE]. Slides: cover, definition, one part each, situation, do-not-use, CTA. Do not invent an extra part. If the situation would need a metric I did not give, write [VERIFY] there rather than inventing proof. Keep the name consistent.
Pro tip: Use a name you already say out loud. Footnote names do not travel in comments.
If-then decision filter
27/30✨ What it does
Puts each real fork on its own IF/THEN slide plus your default. Refused forks stay out, and missing thresholds stay tagged.
Write an if-then carousel for [DECISION]. Forks I actually use: [FORKS]. Default when unsure: [DEFAULT]. Fork I refuse: [OFF LIMITS FORK]. Audience: [AUDIENCE]. Cover, who it is for, one fork per slide as IF headline and THEN body (20 to 30 words), default slide, closer. Honor the refused fork. Do not invent a fork. If a then-clause needs a number I did not give, mark [VERIFY] instead of inventing a threshold. No ASCII flowchart.
Pro tip: One fork per slide. Three ifs on one slide means nobody screenshots it.
Scorecard people can save
28/30✨ What it does
Builds a saveable scorecard from your criteria and good-versus-bad notes. Extra criteria and invented numbers stay out.
Outline a scorecard carousel for [RUBRIC]. Criteria I use: [CRITERIA]. What good versus bad looks like in my words: [GOOD BAD]. Audience: [AUDIENCE]. Cover on the bad buy this prevents, how to use it in 20 words, one criterion per slide, total-score meaning, CTA [CTA]. Do not invent a criterion. Do not invent numbers. If scoring needs a scale I did not set, write [VERIFY] and ask me for the scale. No HTML table.
Pro tip: Score in words they can use on a call. High synergy is not a score.
Teardown of a public page, facts on the page
29/30✨ What it does
Teardown slides each carry a short word-for-word quote from the pasted page. Traffic numbers and secrets not on the page stay out.
Outline a teardown carousel of a page I am allowed to discuss. Page text I pasted: [PASTE PAGE TEXT]. What I will judge: [CRITERIA]. What I will not claim: [OFF LIMITS]. Audience: [AUDIENCE]. Cover, one criterion per slide with a quote from the paste, closer. Keep quotes short and word for word. Do not invent a traffic number, a conversion rate or a brand secret. If a judgment needs a fact not in the paste, write [VERIFY]. No dunking for sport.
Pro tip: Paste the live text. Memory-quotes are how teardowns go unfair.
One framework, three audiences
30/30✨ What it does
Writes three 6-slide mini-outlines that keep part names identical while changing the cover language. Missing examples stay tagged, not borrowed.
Adapt one framework for three audiences without changing the parts. Framework parts: [PARTS]. Audiences: [AUDIENCE A], [AUDIENCE B], [AUDIENCE C]. For each audience write a 6-slide mini-outline: cover in their words, three part-slides, one example slot tagged if I did not give an example, closer. Do not invent an example from another industry. If an example is missing, write [VERIFY] in that slot. Keep part names identical across the three so I can reuse art.
Pro tip: Reuse the part names in art. Renaming parts per audience triples design time.
Free tool
Resume Bullet Rewriter
Turn a weak resume bullet into 3 strong, metric-driven versions.
Frequently Asked Questions
Prompts are the starting line. Tutorials are the finish.
A growing library of hands-on tutorials on ChatGPT, Claude, Midjourney, and many other AI tools. New tutorials are added regularly.
Related guides