ChatGPT prompt for presentations
Copy this prompt, change the first line to your topic, and send it to ChatGPT, Claude or Gemini. Paste the answer into Presenter Cue for finished slides.
Step by step: get better answers from your AI
- Give it facts. The more you put in the topic box (audience, numbers, the point you want to make), the less you have to fix. Ask for the length you actually have time for.
- Ask for changes in plain words. Keep the chat open: “make slide 4 shorter”, “add a slide on pricing”, “simpler words”. Ask it to send the whole JSON again, then paste it in.
- Cut off or messy? Paste it anyway. Presenter Cue repairs common mistakes (smart quotes, trailing commas, a cut-off end) and reads plain outlines too, so you rarely need to ask again.
The prompt
Create a business presentation about: [your topic]
Details:
- Audience: [who will watch]
- Number of slides: 10
- Tone: professional
Return ONLY one JSON code block, with no text before or after it, in exactly this shape:
```json
{
"title": "Deck title",
"theme": "wine",
"slides": [
{ "layout": "title", "title": "…", "subtitle": "…", "notes": "…" },
{ "layout": "bullets", "title": "…", "bullets": ["…", "…", "…"], "notes": "…" }
]
}
```
Layouts you can use, and the fields each one takes:
- "title": the opening slide: deck title and a one-line subtitle. Fields: "title", "subtitle", "image": { "prompt": "photo idea" }
- "section": a divider that starts a new part of the talk. Fields: "kicker" (e.g. "Part 2"), "title", "subtitle"
- "agenda": a numbered list of what the talk covers (3 to 8 items). Fields: "title", "bullets": ["…"]
- "bullets": a heading with 3 to 5 short points. Fields: "title", "subtitle", "bullets": ["…"]
- "two-column": two groups side by side, each with a heading and points. Fields: "title", "left": { "heading": "…", "bullets": ["…"] }, "right": { "heading": "…", "bullets": ["…"] }
- "image-left": a picture on the left, heading and points on the right. Fields: "title", "bullets": ["…"], "image": { "prompt": "photo idea" }
- "image-right": heading and points on the left, a picture on the right. Fields: "title", "bullets": ["…"], "image": { "prompt": "photo idea" }
- "image-full": one full-bleed picture with a caption over it. Fields: "title", "subtitle", "image": { "prompt": "photo idea" }
- "quote": one short quotation and who said it. Fields: "quote", "author"
- "stats": 2 to 4 key numbers, each with a short label. Fields: "title", "stats": [{ "value": "42%", "label": "…" }]
- "chart": a bar, line or pie chart, with up to 3 takeaways beside it. Fields: "title", "chart": { "type": "bar", "labels": ["…"], "values": [1, 2], "unit": "%" }, "bullets": ["…"]
- "timeline": 3 to 6 dated milestones in order. Fields: "title", "steps": [{ "date": "2025" (timeline only), "title": "…", "text": "…" }]
- "process": 3 to 6 numbered steps, each with a short explanation. Fields: "title", "steps": [{ "date": "2025" (timeline only), "title": "…", "text": "…" }]
- "comparison": two options compared: before and after, pros and cons, us and them. Fields: "title", "left": { "heading": "…", "bullets": ["…"] }, "right": { "heading": "…", "bullets": ["…"] }
- "closing": the last slide: thanks, a call to action, contact details. Fields: "title", "subtitle", "bullets": ["…"]
Rules:
- Start with a "title" slide and end with a "closing" slide.
- Use a mix of layouts that fits the content; don’t use "bullets" for every slide.
- Slide titles: at most 8 words. Bullets: at most 5 per slide, each at most 12 words. No full paragraphs on slides.
- "notes" are speaker notes: 2 to 4 natural sentences the presenter can say for that slide. Every slide has notes.
- Charts: "type" is "bar", "line" or "pie"; "labels" and "values" have the same length; values are plain numbers. If you don’t have real figures, make the title say they are example figures.
- For image layouts, describe a fitting photo in "image": { "prompt": "…" }. Don’t include image URLs.
- Never invent quotes from real people; use a quote only if it is well known and correctly attributed.
- Use plain text: no Markdown, emoji or HTML inside the JSON.
- The JSON must be valid: double quotes, no comments, no trailing commas.Easier: open the studio, describe your talk and tap your assistant. The studio fills in this prompt for you.
Why this prompt works
- It asks for structure. Each slide has a layout, so the answer becomes a real deck, not a document.
- It sets limits. Short titles and at most five short bullets keep slides readable from the back of the room.
- It asks for speaker notes. You get something to say, not just something to show.
- It forbids invented quotes and asks for example figures to be labelled, which keeps the deck honest.
Follow-up prompts
- “Make it 6 slides and keep the strongest points.”
- “Rewrite for a non-technical audience.”
- “Add a chart slide with the numbers I give you: …”
- “Send the whole JSON again with these changes.”
The format and rules part, explained
Every Presenter Cue prompt has two parts. The first describes your presentation: the topic, the audience, how many slides and the tone. The second, which starts with “Return ONLY one JSON code block”, is the same in every prompt. It is what makes the answer paste straight into slides. Here is what each piece does:
- “Return ONLY one JSON code block” keeps the assistant from wrapping the deck in an essay. Presenter Cue can still find the JSON if it does, but a clean answer is faster.
- The example shape shows the fields a deck has: a title, a theme and a list of slides, each with a layout, a title and its content. Assistants copy shapes they are shown more reliably than shapes they are told about.
- The list of layouts tells the assistant which slide types exist and which fields each needs: bullets for Bullets, values and labels for Chart, dated steps for Timeline. That is why answers come back with charts and timelines, not just bullet slides.
- Length rules (short titles, a few short bullets) keep slides readable from the back of a room.
- Speaker notes ask for sentences you can say aloud, so the screen stays clean.
- “Don’t invent facts” asks for [placeholders] where a figure is missing, which makes gaps easy to find.
- Valid JSON rules (double quotes, no trailing commas) prevent the small syntax errors assistants sometimes make. Presenter Cue repairs most of those anyway.
Changing the first part
Everything above “Return ONLY” is yours to edit: the topic, the details, the audience. Add facts as a list, say what to emphasise, name sections you want. Leave the second part as it is unless you know you want something different; for example, you can delete the notes rule if you don’t want speaker notes.
Using it outside Presenter Cue
The prompt works in any assistant that can follow instructions: ChatGPT, Claude, Gemini, Copilot and others. The JSON it returns is a plain, readable description of a deck, so even without Presenter Cue you can read it as an outline. With Presenter Cue, it becomes slides you can edit, present and download.
Questions
Can I use this prompt without Presenter Cue?
Yes, but the answer is JSON meant for a slide renderer. Presenter Cue is free and turns it into slides in a few seconds.
What if ChatGPT ignores the format?
Ask: “Send it again as one JSON code block in the shape I gave you.” Presenter Cue also reads plain outlines as a fallback.
Why JSON instead of plain text?
JSON keeps each slide’s parts separate: the layout, the title, the bullets, the chart values, the notes. That is what lets Presenter Cue build real charts and timelines rather than pasting text into boxes.
Can I remove the speaker notes from the prompt?
Yes. Untick Speaker notes in the studio, or delete that rule from a copied prompt. Slides then come back without notes.
Does the prompt work in other languages?
Yes. Add “Write every slide and note in [language]” to the details, or set the Language field in the studio, which adds that line for you.
What if the assistant changes the layout names?
Presenter Cue understands common alternatives, such as “bullet points” for Bullets or “metrics” for Big numbers, and falls back to Bullets for anything it doesn’t recognise.
Last updated .