ChatGPT Prompts for PowerPoint Presentations

Quick-Start (Copy This Right Now)

A sales director once walked into a client pitch with ChatGPT-generated slides so text-heavy that halfway through, he noticed the deal team had stopped reading along and started skimming ahead, waiting for him to catch up to slides they'd already finished. That's the single most common failure with chatgpt presentation prompts: asking for slide content the way you'd ask for an essay, and getting paragraphs where you needed a handful of words per slide.

Create an outline for a 10-slide presentation on [topic]. For each 
slide, give me: a slide title, 3-4 short bullet points (each under 10 
words), and separately, longer speaker notes with the full detail I'll 
actually say out loud.

What this does: separating brief on-slide bullets from detailed speaker notes is the single change that fixes the text-wall problem -- the audience sees a scannable slide, and you have the full detail to actually speak from, instead of both living in the same overloaded text block.

⚡ Pro tip: Always request slide bullets and speaker notes as two separate outputs, with an explicit word limit on the bullets. This one instruction prevents the most common and most damaging presentation mistake: slides so dense the audience reads instead of listens.

Understanding the Variables

The variable that matters most for presentation prompts is audience and purpose -- a board update, a sales pitch, and a training session all need fundamentally different slide density, tone, and structure, even covering similar underlying content. A generic "make me slides about X" prompt has no way to calibrate for these differences.

⚠️ Common mistake: Requesting presentation content without specifying the audience or purpose. The same topic needs completely different slide structures for a 5-minute investor pitch versus a 45-minute internal training session, and a prompt that doesn't specify which one produces content miscalibrated for whichever context you actually need.

Step-by-Step: Building a Presentation

Start with the structural outline, specifying audience and time constraint explicitly:

Outline a 15-minute presentation for a board update. Audience: board 
members who know the company well but haven't seen this quarter's 
detailed numbers. Purpose: get approval for a budget increase in the 
marketing department. Structure should build toward the ask, not bury 
it at the end.

What this does: specifying that the structure should build toward the ask, rather than defaulting to a chronological "here's what happened this quarter" structure, keeps the presentation purposeful -- every slide leads somewhere, rather than existing as background before the actual request finally shows up on slide 12.

A nonprofit program director uses a similar structural approach for grant presentations, where the funder's specific priorities need to shape the slide order:

Outline a 10-slide presentation for a grant funder whose stated 
priority is measurable community impact, not organizational history. 
Structure so impact data and outcomes appear in the first 3 slides, 
not at the end after background information.

What this does: front-loading the content the specific audience cares most about, rather than a generic chronological narrative, respects that funders reviewing many proposals often decide their level of attention in the first few slides -- burying the strongest content late risks losing their attention before they get there.

Pro-Level Variations

For slides with data, the challenge shifts from wording to deciding what belongs on the slide versus what belongs only in speaker notes. A financial analyst preparing quarterly results slides uses ChatGPT to help make that call explicitly:

I have this data: [paste table]. Tell me which 2-3 numbers are most 
important to show directly on the slide (as a simple chart or big 
number callout), and which details should move to speaker notes 
instead of cluttering the slide itself.

What this does: explicitly asking which numbers matter most, rather than trying to fit the whole table onto a slide, forces the same slide-versus-notes separation that fixes text-heavy slides, applied specifically to data rather than prose.

⚠️ Common mistake: Putting an entire data table on a slide because "it's all relevant," rather than selecting the 2-3 numbers that actually matter for this particular audience and leaving the rest in speaker notes or an appendix. A crowded data slide gets skimmed, not read, defeating the purpose of showing the data at all.

A trainer building an onboarding presentation for new hires uses ChatGPT to adjust slide density specifically for a training context, where some repetition and detail actually helps retention rather than hurting it:

For this specific training presentation (not a pitch or update), 
slightly more text per slide is appropriate since attendees will 
reference these slides again later. Still keep bullets scannable, but 
you can use slightly longer phrases than you would for a live pitch 
deck.

What this does: explicitly recalibrating the density guidance for a reference document rather than a live-presented pitch acknowledges that the "keep it minimal" rule isn't universal -- a training deck that gets referenced later benefits from more self-contained detail than a deck meant to be spoken alongside.

⚡ Pro tip: Don't apply the same minimal-text rule to every presentation type. Live pitch decks need scannable brevity because you're speaking alongside them; reference materials and training decks that stand alone need more self-contained detail. State which type you're building so the density calibrates correctly.

Troubleshooting Common Issues

If slides keep coming back too text-heavy despite asking for bullets, the fix is usually to add an explicit word count ceiling per bullet, not just "keep it brief." Vague brevity instructions get interpreted more loosely than a hard number like "under 8 words per bullet."

Visual and Design Suggestions

Beyond text content, ChatGPT can also help think through what a slide should visually emphasize, even without generating the actual graphic. A product manager preparing a roadmap presentation uses this to plan visual hierarchy before building the slide:

For a slide showing our product roadmap across 4 quarters, suggest a 
visual layout: should this be a horizontal timeline, a 4-column grid, 
or something else? Consider that the audience needs to quickly see 
which quarter has the most planned work.

What this does: asking for a layout recommendation based on what the audience needs to quickly grasp (workload distribution across quarters) rather than a generic "make it look nice" request produces a suggestion tied to actual communication goals, which is what should drive slide design decisions in the first place.

⚡ Pro tip: When asking for visual layout suggestions, always state what you want the audience to understand at a glance. The right layout depends entirely on the specific insight you're trying to convey, not on general design principles alone.

A startup founder building an investor deck uses ChatGPT to stress-test whether a slide's core message is actually clear, a check that's easy to skip when you're too close to your own content:

Here's the speaker notes content for my "market opportunity" slide: 
[paste]. If someone only glanced at the slide for 5 seconds without 
hearing me speak, what's the one thing they'd walk away understanding? 
Is that the actual point I need this slide to make?

What this does: simulating the 5-second glance test surfaces whether a slide's headline message actually lands without narration, which is a useful check since investors flipping through a deck later (without you presenting) will experience it exactly that way.

⚠️ Common mistake: Assuming a slide is clear because it makes sense when you're presenting and narrating alongside it. Many decks get reviewed later without the presenter in the room -- testing whether a slide communicates its core point without narration catches gaps that only show up in that standalone context.

A conference speaker preparing a technical talk uses ChatGPT to plan pacing across an entire deck, checking that slide density doesn't cluster unevenly across the talk's time allocation:

I have 20 slides for a 25-minute talk. Here's a one-line summary of 
each slide's content: [list]. Flag if any section seems to have too 
many slides for too little time, or too few slides for a section that 
needs more depth, based on typical pacing of about 1 slide per minute.

What this does: checking the distribution of slides against a rough pacing benchmark catches structural pacing problems (a rushed ending crammed into 3 slides after a slow, over-explained opening) before they become a live timing problem during the actual talk.

Your Turn

Take a presentation you're currently building and try the two-part structure -- brief bullets plus separate detailed speaker notes -- even if you've already drafted it in one dense block. Notice how much more scannable the slide content becomes once it's not trying to hold both the summary and the full explanation at once.

If you're building recurring presentations -- quarterly updates, pitch decks, training materials -- it's worth saving the audience-specific structure and density rules for each type, so you're not recalibrating from scratch every time. PromptABCD works well for storing these presentation templates by type, ready to reuse the next time a similar deck comes up.