Gemini Prompts for Presentations

A product manager once used Gemini to generate slide content for a leadership pitch, pasted the output directly into a deck without restructuring it, and presented eleven slides of dense bullet points to a leadership team that had asked specifically for a five-minute pitch. He got cut off at slide four. The problem wasn't the content quality — it was that the prompt asked for "slide content" without any constraint on how much a single slide should actually hold, which is the single most common failure in gemini presentation prompts.

Before: The Weak Prompt

Create slide content for a presentation about our Q3 marketing results.

What this does (poorly): produces reasonable-sounding content with no constraint on density, length, or the actual time available to present it, so Gemini defaults to comprehensive coverage — thorough, accurate, and completely unusable for a time-boxed presentation, since comprehensive and presentable are different goals that this prompt conflates.

Why It Fails

Presentations fail on density far more often than they fail on content quality. A slide with six bullet points, each a full sentence, isn't wrong information — it's a document pretending to be a slide, and it forces the audience to choose between reading it or listening to the presenter, since doing both simultaneously isn't actually possible. This is such a common failure that "death by bullet points" is a cliché for a reason, and it happens because most content-generation requests never specify a length or density constraint tied to how the content will actually be delivered.

⚠️ Common mistake: Asking for "slide content" without specifying how much time will be spent on each slide. A prompt with no pacing constraint will produce content sized for a document, not a timed spoken presentation, and the mismatch only becomes obvious once you're standing in front of the actual audience.

After: The Improved Prompt

I have 5 minutes to present Q3 marketing results to a leadership team that wants outcomes, not process details. Here's the raw data: [PASTE KEY METRICS AND CONTEXT]

Create content for exactly 4 slides:
Slide 1: One headline result (the single most important number), one supporting sentence
Slide 2: What drove that result — maximum 3 bullet points, each under 10 words
Slide 3: One specific risk or challenge, stated directly, with the mitigation in one sentence
Slide 4: The specific ask or decision needed from this audience, stated as a direct question

For each slide, write the actual spoken talking points separately from the on-screen text — the screen should show far less than what I'll actually say.

What this does: builds the time constraint directly into the structure (4 slides for 5 minutes is a realistic pace) rather than generating unconstrained content and hoping it fits, and the instruction to separate spoken talking points from on-screen text fixes the core "document pretending to be a slide" problem by explicitly allowing the screen to show less than the presenter says.

⚡ Pro tip: Always specify your actual time limit and audience expectation in the prompt itself, not just the topic. "5 minutes to a leadership team that wants outcomes, not process" produces fundamentally different, more usable content than "presentation about marketing results," because it constrains both length and what kind of information actually matters to include.

Breaking Down Each Element

The screen-versus-spoken split is the single most valuable instruction in this whole prompt. Most people default to writing slides that contain everything they plan to say, which forces the audience to read instead of listen. Separating the two forces genuinely sparse on-screen text — often just the headline number or a short phrase — while the fuller explanation lives in what the presenter actually says aloud, which is how experienced presenters naturally work but rarely how AI-generated slide content defaults to working without an explicit instruction.

The "stated as a direct question" instruction for the final slide matters because a vague closing slide ("thoughts and discussion") produces exactly the kind of ambiguous ending that lets a meeting drift without a decision, while a direct question ("should we reallocate the Q4 budget toward the channel that outperformed") forces the actual decision point the presentation exists to reach.

A director of sales operations at a manufacturing company uses this exact structure for weekly leadership updates and specifically credits the direct-question closing slide with cutting the average length of the discussion portion of these meetings, since leadership no longer has to figure out what decision is actually being requested — it's stated plainly on the last slide.

⚡ Pro tip: For any high-stakes presentation, generate the spoken talking points first and read them aloud with a timer before finalizing the slides. If the talking points alone run over your time limit, no amount of slide design will fix the actual problem, which is too much content for the available time.

Variations for Different Contexts

For a longer, more detailed internal training presentation where audience engagement rather than time pressure is the main constraint, the density rules relax somewhat, but the screen-versus-spoken separation still applies — even a 45-minute training session benefits from slides that support the spoken explanation rather than duplicating it verbatim.

For an investor pitch deck, a founder at an early-stage logistics startup adapts the structure to fit a stricter narrative arc (problem, solution, traction, ask) with an explicit instruction that each slide must connect logically to the next, since investor pitches specifically fail when they read as a list of disconnected facts rather than a coherent argument building toward the ask.

⚠️ Common mistake: Applying the same density rules to a presentation that will be sent as a stand-alone document (read without a live presenter) as to one that will be presented live. A deck meant to be read on its own needs more self-contained on-screen text, since there's no spoken explanation filling the gaps — using the live-presentation density rules for a document-style deck leaves readers with sparse, confusing fragments.

Save and Reuse This

The reusable structure here isn't any specific slide content — it's the constraint-setting habit: specify actual time available, specify audience expectations, separate spoken content from on-screen text, and end with a direct question rather than a vague discussion prompt. New topic, new data, same structural discipline underneath.

Once you've built a presentation prompt structure that consistently produces decks people can actually present within their time limit, keep it accessible for your whole team rather than relearned the hard way by each new presenter. PromptABCD gives teams a shared, versioned prompt library for exactly this, so the four-slide, five-minute structure that saved one presenter from getting cut off mid-pitch becomes the default the next person reaches for, instead of everyone rediscovering the density problem independently in front of their own leadership team.

More Scenarios Worth Studying

A conference speaker preparing a 20-minute technical talk for a developer audience uses a time-boxed version of this structure broken into segments — roughly one slide per 90 seconds of planned speaking time — and specifically found that forcing herself to write the spoken talking points first, before touching slide design, revealed that her original outline had nearly twice as much content as would actually fit, well before she ever stood in front of an audience with a countdown timer.

A nonprofit program director preparing a grant renewal presentation to a funding committee uses the direct-question closing slide adaptation specifically for the funding ask, replacing a vague "we hope you'll consider continued support" closer with a specific, stated request tied to a specific dollar amount and program outcome — a change she credits with making the committee's actual decision process faster, since the ask was unambiguous rather than something committee members had to infer from surrounding context.

A middle manager delivering a reorg announcement to her team uses a modified version of the structure adapted for a harder emotional context, where "outcomes not process" doesn't quite apply — she specifically asks for content that states the change directly on slide one rather than building up to it, since burying difficult news in a delayed reveal (a common instinct when the news is unwelcome) tends to increase anxiety in the room rather than reduce it, as people sense something is coming before it's actually said.

A Note on Design Versus Structure

None of the structural fixes above address visual design — font choices, color schemes, slide layout — which is a separate skill from the content and pacing discipline covered here. A well-structured, appropriately sparse slide can still look unprofessional if the visual design is inconsistent or hard to read, and a beautifully designed slide can still fail if it's packed with too much content for the time available. Both matter, but they're solved differently: content structure through prompts like the ones above, visual design through a template or design tool applied consistently across the deck. Confusing the two — assuming a content fix will also fix a design problem, or vice versa — is a common source of frustration when a deck still doesn't land despite one half of the problem being solved.