slug: headline-test view: skill layer: slide agent: architect audience: llm companion: corpus/storymakers/frameworks/slide/headline-test.md

Headline Test — operational reference

Compact reference for an LLM checking or generating slide titles and any line whose job is to summarise what follows. Showcase version (headline-test.md) explains the why. This is the chuleta.

One sentence. A title passes if and only if it has a real verb, makes a defensible claim, and would leave the reader with the right point if they read nothing else.

Decision tree

Q1. Does the title contain a real verb?
    (NOT "is", "has", "are", "was", "have", "had" — copulas don't count)
    NO  → REJECT. Rewrite with a working verb. STOP.
    YES → Q2

Q2. Does the title assert a claim a reasonable reader could disagree with?
    NO  (it's a tautology, a topic, or a meta-description) → REJECT. Rewrite. STOP.
    YES → Q3

Q3. If the audience read only this line, would they leave with the slide's main point?
    NO  → REJECT. The title decorates rather than argues. Rewrite. STOP.
    YES → PASS. Title is action-grade.

Recipe (apply → fail any → regenerate → repeat)

For every candidate title, in order:

Read it aloud. If it doesn't sound like a sentence, it's a label — fails Q1 immediately.
Strike every copula (is, has, are, was, have, had). What verb is left? If none, fails Q1. Stop and rewrite.
State the inverse. If the inverse is incoherent or trivially uninteresting, the title is a tautology — fails Q2. Stop and rewrite.
Cover the slide. Read only the title. Does the reader leave with the slide's main point? If not, fails Q3 (the what-without-so what failure). Stop and rewrite.
Loop. Regenerate, retest from Q1. Don't ship a failed title; don't paper over a failure with longer wording.

Anti-patterns (reject on sight)

Pattern	Why it's wrong
"Margins" / "Cost dynamics" / "Strategy"	Label, not title. No verb, no claim.
"Margins are under pressure"	Copula as only verb — existence, not finding.
"Margins have changed over time"	Verb-bearing tautology. Inverse is incoherent.
"Margins may have somewhat softened"	Hedged into nothingness. No defensible claim.
"This slide shows the margin trajectory"	Meta-description. Passes verb test, fails memorability.
"Margins have been impacted by several factors"	Passive voice hiding agent. Force by what?
"Phase 2: Profitability analysis"	Navigation token. Fine on a divider, wrong on a body slide.

When NOT to use it

Almost universally applicable. Narrow exceptions: divider / section slides (navigation, not summary), agenda slides (intentionally topic-shaped), and neutral-framing research presentations where the title shouldn't pre-judge data. In all three the title is navigation, not argument — different rule.

Output checklist

A title is shippable when:

Contains a verb that does work (not a copula).
Asserts a claim with a defensible inverse.
Carries the slide's main point on its own.
Reads aloud at conversational pace in one breath.
Specifies the so what when a number is involved.
No hedging adverbs unless uncertainty is the finding.
No passive voice unless the agent is genuinely unknown.

Canonical signals (for matchers)

Keywords: headline, subhead, action title, slide title, governing thought, kicker, lede, topic title, so what.
Component kinds: title (primary — action-title pattern), subtitle, callout (when callout is the apex statement).
Slide-level shape: title is a declarative sentence with a working verb whose body proves the sentence. A noun-phrase title or copula-only title is the failure mode this test exists to catch.
Commentary signal: prose discussing whether a headline "lands", "earns the slide", or "passes the verb test" invokes this canon.

References

AP Stylebook — active verbs in headlines; rule against label heads.
Zelazny, Say It with Charts, 1985 — action-title doctrine for the deck.
Minto, The Pyramid Principle, 1973 — declarative titles at every node.
Clark, Writing Tools, 2006 — Tool 3, Activate your verbs (Poynter).
Crestodina, Orbit Media — How to Write Truly Great Headlines.
Nielsen Norman Group, 2006 / 2017 — F-pattern: titles carry disproportionate attention.