Speech Writing
Parent: Writing and Documentation · Published reference · snapshot 2026-09-08 · skill executive-comms/references/speech-writing.md
↓ Facts as markdownall context files
A speech is built to be heard, not read. The writer's job is to construct sentences that survive the air between the speaker's mouth and the audience's ear — sentences that can be re-listened-to in re
These notes link each claim to its source. Some citations identify archived research reports that are no longer hosted on this site. A published reference means the content is available; it does not certify independent review or accuracy.Read the editorial policy and follow the sources before relying on a claim.
Overview
- A speech is built to be heard, not read. The writer's job is to construct sentences that survive the air between the speaker's mouth and the audience's ear - sentences that can be re-listened-to in real time only by the listener's short-term memory. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#overview]
1. The talk arc (Duarte / Gallo synthesis)
- Every well-formed speech has the same five-beat skeleton: [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#1-the-talk-arc-duarte-gallo-synthesis]
- Hook - a sentence, image, statistic, or question that earns the next 30 seconds. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#1-the-talk-arc-duarte-gallo-synthesis]
- Frame - what this talk is about and why the audience should care. One sentence. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#1-the-talk-arc-duarte-gallo-synthesis]
- Three acts - the body. The rule of three is not aesthetic; it is cognitive. Three things stay with the audience, five don't. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#1-the-talk-arc-duarte-gallo-synthesis]
- Turn - the moment where the talk pivots from explanation to implication. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#1-the-talk-arc-duarte-gallo-synthesis]
- Call to action - what the audience should do, feel, or remember. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#1-the-talk-arc-duarte-gallo-synthesis]
2. The 18-minute rule
- Carmine Gallo's distillation of TED's curation rule: 18 minutes is the ceiling for a focused, memorable single-speaker talk. For longer slots, structure as multiple sub-talks with clear act breaks. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#2-the-18-minute-rule]
3. The rule of three
- Three is the smallest number that creates a pattern. Use it for: [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#3-the-rule-of-three]
- Top-level structure (three acts, three stories, three asks) [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#3-the-rule-of-three]
- Sentence-level lists ("government of the people, by the people, for the people") [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#3-the-rule-of-three]
- Repeated refrains (Jobs: "Stay hungry. Stay foolish." - repeated three times) [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#3-the-rule-of-three]
4. Applause-line placement and craft
- Place applause lines at act transitions and the CTA. Never two in a row. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#4-applause-line-placement-and-craft]
5. Speech reads differently than essay
- Read every draft aloud. If you run out of breath, the sentence is too long. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#5-speech-reads-differently-than-essay]
6. Stage-direction annotation craft
- Stage directions are inline notes to the speaker about delivery. They live in the delivery script, not the published transcript. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#6-stage-direction-annotation-craft]
- [pause] - short pause, one breath [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#6-stage-direction-annotation-craft]
- [long pause] - full silence, 2–4 seconds [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#6-stage-direction-annotation-craft]
- [lower voice] - drop volume to draw the audience in [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#6-stage-direction-annotation-craft]
- [slow] - drop pace to weight the line [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#6-stage-direction-annotation-craft]
- [wait for laughter] - explicitly note that the speaker should hold [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#6-stage-direction-annotation-craft]
- Place directions on their own line above the line they modify, in brackets. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#6-stage-direction-annotation-craft]
7. Trust the audience
- Amateur speeches over-explain. Professional speeches withhold, then resolve. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#7-trust-the-audience]
- Tell a story before naming its lesson; let the room reach the conclusion one beat ahead. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#7-trust-the-audience]
- Use specific detail, not abstraction. "A blue notebook on the kitchen table" beats "a personal object." [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#7-trust-the-audience]
- Cut your second-best example. Two strong points land harder than three uneven ones. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#7-trust-the-audience]
- Don't apologize for the topic, the time slot, or the room. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#7-trust-the-audience]
8. Ghostwriting ethics
- Capture their voice. Listen to 30+ minutes of their existing speech before drafting. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#8-ghostwriting-ethics]
- Their stories, not yours. Use anecdotes from their life. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#8-ghostwriting-ethics]
- No misattributed lived experience. Never invent a personal anecdote they did not actually experience. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#8-ghostwriting-ethics]
- They get final cut. The speaker must be able to change anything. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#8-ghostwriting-ethics]
Anti-Patterns
- The written-essay-read-aloud. Long sentences, semicolons, nested clauses. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#anti-patterns]
- The five-point list. Audiences forget points four and five. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#anti-patterns]
- The apology open. "I'm not really a speaker…" [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#anti-patterns]
- The unmotivated joke. Humor that doesn't serve the arc. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#anti-patterns]
- The buried CTA. Telling the audience what to do in the third-to-last sentence. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#anti-patterns]
- Stage directions in the published transcript. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#anti-patterns]
References
- Nancy Duarte, Resonate (Wiley, 2010). [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#references]
- Carmine Gallo, Talk Like TED (St. Martin's Press, 2014). [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#references]
- Steve Jobs, Stanford Commencement Address, June 12, 2005. [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#references]
- Sam Leith, Words Like Loaded Pistols (Basic Books, 2012). [archived source: https://llms-explorer.com/sources/mdb-context-hub/speech-writing/#references]
Where this helps
- Structuring a keynote or conference talk around the five-beat talk arc (hook, frame, and the rest) instead of a loose list of points. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Deciding how long a talk should run — the 18-minute rule as a ceiling for a single focused talk, restructured into sub-talks with act breaks for longer slots. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Ghostwriting a speech for an executive, where capturing their actual voice and using only their real anecdotes matters more than polished prose. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Annotating a delivery script with stage directions (pause lengths, emphasis) that live in the speaker's copy but never appear in the published transcript. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
How to apply this
- Read every draft aloud during revision; running out of breath mid-sentence is the tell that a sentence is too long for spoken delivery. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Use the rule of three deliberately — at the structural level (three acts, three stories) and the sentence level — because three is the smallest number that reads as a pattern. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Place applause lines only at act transitions and the call to action, never two in a row, so each one still lands. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Withhold the lesson of a story until after the story is told, letting the audience arrive at the conclusion a beat ahead of the speaker naming it. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Common mistakes
- Writing the written-essay-read-aloud: long sentences, semicolons, and nested clauses that work on the page but collapse when spoken. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Using the five-point list structure, which audiences reliably forget past point three or four. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Opening with an apology ("I'm not really a speaker...") instead of the hook. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Explaining the point of a story before telling it, over-explaining instead of trusting the audience to follow. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Limitations
- The 18-minute rule is a curation heuristic (TED's), not a law — some formats and audiences tolerate or expect longer single-speaker talks, and the rule needs adapting rather than applying blindly. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Ghostwriting ethics require capturing a speaker's real anecdotes and voice, which means a ghostwriter cannot invent a personal experience the speaker didn't actually have, no matter how well it would fit the arc. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Ghostwriting quality depends on access — the guidance recommends 30+ minutes of listening to the speaker's existing material before drafting, which isn't always available on tight timelines. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Stage directions belong only in the delivery script; conflating them with the published transcript confuses the two very different audiences — the speaker delivering it live, and readers encountering it after the fact. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Where this helps
- Structuring a keynote or conference talk around the five-beat talk arc (hook, frame, and the rest) instead of a loose list of points. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Deciding how long a talk should run — the 18-minute rule as a ceiling for a single focused talk, restructured into sub-talks with act breaks for longer slots. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Ghostwriting a speech for an executive, where capturing their actual voice and using only their real anecdotes matters more than polished prose. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Annotating a delivery script with stage directions (pause lengths, emphasis) that live in the speaker's copy but never appear in the published transcript. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
How to apply this
- Read every draft aloud during revision; running out of breath mid-sentence is the tell that a sentence is too long for spoken delivery. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Use the rule of three deliberately — at the structural level (three acts, three stories) and the sentence level — because three is the smallest number that reads as a pattern. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Place applause lines only at act transitions and the call to action, never two in a row, so each one still lands. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Withhold the lesson of a story until after the story is told, letting the audience arrive at the conclusion a beat ahead of the speaker naming it. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Common mistakes
- Writing the written-essay-read-aloud: long sentences, semicolons, and nested clauses that work on the page but collapse when spoken. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Using the five-point list structure, which audiences reliably forget past point three or four. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Opening with an apology ("I'm not really a speaker...") instead of the hook. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Explaining the point of a story before telling it, over-explaining instead of trusting the audience to follow. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Limitations
- The 18-minute rule is a curation heuristic (TED's), not a law — some formats and audiences tolerate or expect longer single-speaker talks, and the rule needs adapting rather than applying blindly. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Ghostwriting ethics require capturing a speaker's real anecdotes and voice, which means a ghostwriter cannot invent a personal experience the speaker didn't actually have, no matter how well it would fit the arc. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Ghostwriting quality depends on access — the guidance recommends 30+ minutes of listening to the speaker's existing material before drafting, which isn't always available on tight timelines. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
- Stage directions belong only in the delivery script; conflating them with the published transcript confuses the two very different audiences — the speaker delivering it live, and readers encountering it after the fact. [source] — AI-suggested, synthesized from this pack's existing facts — not extracted from a source document.
Children
- No children recorded.