The Best Prompting Advice Came From a Writing Teacher
Ernie Svenson · email · 2026-04-21 Reader 🖥️
The biggest barrier to getting useful output from AI is not prompting technique but the same problem William Zinsser identified in bad writing — people try to communicate before they have figured out what they actually want to say; therefore the single most leveraged improvement most users can make is to pause and articulate their intent in one clear sentence before they prompt.
practical · argumentative — A short practical essay aimed at lawyers and other professionals struggling with AI tools. It argues that William Zinsser's clear-writing advice from On Writing Well applies directly to prompting AI. The form is argumentative-instructional: it diagnoses a problem (vague prompts), names a cause (unclear thinking), and prescribes a remedy (Zinsser's question — what am I trying to say?). Light on technique, heavy on diagnosis and motivation.
Structure
| Part | Role |
|---|---|
| Zinsser hook: 'Writers must constantly ask: what am I trying to say? Surprisingly often they don't know.' | Establishes the borrowed authority and the central insight that writing problems are thinking problems |
| Diagnosis: 'The real problem isn't the tool' | Reframes user complaints about ChatGPT/Claude — vague prompts produce vague output; the tool is following the user's lead |
| Zinsser's insight applies directly | Explicitly maps the writing-as-thinking diagnosis onto prompting-as-thinking |
| What 'knowing what you want' looks like — purpose, audience, tone, success criteria | Operationalizes the abstract advice into four concrete questions to ask before prompting |
| A simple experiment: pause 10 seconds and answer Zinsser's question | Provides a low-friction practice the reader can adopt immediately |
| Bottom line: figure out what you're trying to say before you try to say it | Restates the thesis and elevates clarity-of-intent above any technique |
Key terms
- prompting Prompting as Communication
The act of instructing an AI tool to produce a specific output — treated by Svenson as a communication act that depends primarily on the user's clarity of intent, not on syntactic technique - knowing what you want Prompting as Communication
Having thought through purpose, audience, tone, and success criteria — at least roughly — before issuing a prompt. Does not require a formal outline, but does require honest internal articulation. - vague prompt / vague result Prompting as Communication
A prompt that fails to specify purpose, audience, tone, or success criteria — and the correspondingly unfocused output the AI returns. The pairing is treated as causal: vague input deterministically produces vague output. - Zinsser's question Prompting as Communication
'What am I trying to say?' — the one-sentence test Svenson borrows from On Writing Well and prescribes as a 10-second pause before prompting - technique Prompting as Communication
Prompt-engineering tricks — formats, structures, special instructions — which Svenson concedes have their place but explicitly subordinates to clarity of intent
Propositions
- explicitThe dominant failure mode in AI prompting is not technical but cognitive — users do not know what they want before they ask.conf: high
- explicitZinsser's diagnosis of bad writing — that it is a thinking problem, not a craft problem — applies directly to bad prompting.conf: high
- explicitVague prompts deterministically produce vague results; the AI faithfully reflects the specificity of the input.conf: high
- explicitcorroboratedKnowing what you want requires at least rough internal answers to four questions: purpose, audience, tone, and success criteria.conf: highsupports ←Prompting Claude Fable 5 · p4 — The Fable 5 'give the reason' template ('I'm working on [larger task] for [who]... they need [what the output enables]') operationalizes exactly Zinsser's pre-prompt questions of purpose and audience — as concrete, model-specific guidance for an agent rather than a writing heuristic.
- explicitIf you cannot describe what a good result looks like, you cannot evaluate the AI's output and cannot guide it toward something better.conf: high
- explicitBeing able to state your intent in one clear sentence is a sufficient readiness test for prompting.conf: high
- explicitcorroboratedClarity-of-intent yields larger improvements in output quality than any specific prompting technique.conf: highsupports ←How Coders Can Survive—and Thrive—in a ChatGPT World · p7 — Abraham's 2023 advice that precision and clarity about what you want drives output quality more than anything else matches the later, more fully argued claim that clarity-of-intent yields larger improvements in AI output than any specific prompting technique.extends ←The habit that makes AI worth using · p3 — Both pieces, from the same author, devalue prompting technique in favor of a more fundamental human skill as the real driver of AI value — clarity of intent in the earlier piece, attention-management plus evaluation-of-information here — extending the same 'it's not the prompt trick' thesis to a new use case (fact-checking claims rather than generating text).
- explicitdevelopedAI tools cannot want something for you — supplying intent remains an irreducibly human job.conf: highextends ←Software Teaming in the Age of AI with James Herr and Woody Zuill · p15 — 'AI tools cannot want something for you — supplying intent remains an irreducibly human job' is the same claim from the prompting-craft side that Woody's taxi-driver / high-level-intent analogy makes from the team-programming side; this episode adds the specific claim that mob-programming discipline is training ground for that skill.
- implicitPrompting is fundamentally a writing act — a communication problem governed by the same rules of clarity that govern good prose.conf: high
- implicitdevelopedThe act of articulating intent before prompting is itself a form of thinking — writing-to-think rather than writing-to-output.conf: highextends ←Writing vs. Thinking · p11 — Zinsser/Svenson extend writing-as-thinking into prompting ("writing-to-think rather than writing-to-output"); this doc extends the same thesis into software naming/design — a parallel domain application of the same underlying claim.
- implicitModels are accurate mirrors of input specificity — they do not actively probe, clarify, or fill in missing intent.conf: medium
- implicitThe professional reader (lawyers in Svenson's audience) is the primary locus of value — the AI is a tool that magnifies, but does not replace, professional judgment about purpose and audience.conf: high
Arguments
The Zinsser principle ('know what you're trying to say') is the single most important prompting advice. strength: medium
Explicit premises:
- Most people who complain about AI output gave vague prompts
- Vague prompts produce vague results because the AI follows the user's lead
- Zinsser observed the same pattern in bad writing fifty years ago — the problem is upstream of craft
- Clarity of intent is achievable through a brief pause and the one-sentence test
Implicit premises:
- Prompting is sufficiently analogous to writing that writing-craft principles transfer cleanly
- Models behave as faithful executors of stated intent rather than as collaborators who interrogate vague requests
- The marginal value of clarity-of-intent exceeds the marginal value of prompting technique for most users at their current skill level
Inferences:
- Teaching prompting should emphasize pre-prompt thinking over prompt-formatting tricks
- Tools or rituals that force users to articulate intent (templates, pause prompts, one-sentence summaries) would yield more improvement than 'better prompts' libraries
Prompting is best understood as a thinking-and-writing problem, not a technical skill. strength: medium
Explicit premises:
- Zinsser's writing advice transfers directly to prompting
- The four pre-prompt questions (purpose, audience, tone, success) are classic communication-and-rhetoric concerns
Implicit premises:
- Whatever is distinctive about LLMs (token sampling, context windows, training data, model-specific quirks) is less consequential than the user's clarity of intent
- The user's job ends at articulating intent; the model's job is to faithfully execute it
Inferences:
- Investment in writing instruction will yield prompting gains
- Communication-trained professionals (lawyers, writers, teachers) should adopt AI faster and better than their less-articulate peers
AI cannot supply intent; humans must. strength: medium
Explicit premises:
- AI tools are powerful but cannot want something for you
- The AI faithfully reflects whatever specificity (or lack thereof) the user provides
Implicit premises:
- Wanting is intrinsically tied to having goals, stakes, and embeddedness in a context that AI lacks
- Models will not (or should not) ask clarifying questions to surface intent the user has not articulated
Inferences:
- The intent-articulation step remains a durable human contribution even as AI capability scales
Narratives
- The restaurant analogy: walking into a restaurant and saying 'bring me some food' will get you something, but you probably won't love it. Vague prompts work the same way.
→ Makes the abstract claim about specificity concrete and intuitive; positions the user as a diner whose dissatisfaction is self-inflicted - The unnamed lawyer who said 'I asked it to write a letter to my client' or 'I told it to help me with a contract provision' — the prototypical vague prompter whose complaint Svenson disarms by asking 'What exactly did you ask for?'
→ Anchors the essay in professional reality and gives the audience a recognizable mirror; the move from complaint to diagnostic question models the intervention Svenson recommends - Zinsser as the borrowed authority: a writing teacher whose million-copy book contains a single line ('Writers must therefore constantly ask: what am I trying to say? Surprisingly often they don't know.') that 'perfectly explains' AI-tool friction.
→ Lends weight via cross-domain authority — a non-AI writing teacher's insight reframes a contemporary technology problem, suggesting the answer was already known and just needed translating
Questions addressed
- ✅ Why do so many people get disappointing results from AI tools?
Because they prompt before they have figured out what they actually want; vague intent produces vague output. - ✅ What is the single highest-leverage improvement most prompters can make?
Pause before prompting and answer Zinsser's question — 'What am I trying to say?' — in one clear sentence. This dominates any technique-level improvement. - ✅ What does it mean, operationally, to 'know what you want'?
Have at least rough internal answers to four questions: purpose, audience, tone, success criteria — even if informal. - ✅ Does this mean prompting technique doesn't matter?
Technique has its place but is downstream of and smaller than clarity of intent for most users. - ⭕ When does the writing-as-prompting analogy break down?
Not addressed. The essay treats the analogy as essentially complete and does not engage with cases where prompting diverges from writing — agentic multi-step work, context engineering, tool use, model-specific quirks, iterative refinement loops. - ⭕ What should you do when you genuinely don't know what you want and are using AI to help discover it?
Unaddressed — Svenson's framing presumes intent precedes prompting, but exploratory prompting (using the AI as a thinking partner to find out what you want) is a major and legitimate use case excluded by the essay's frame.
Gaps
- incompleteThe essay treats the AI as a passive instrument that 'follows your lead.' This was largely true for early single-shot text generation but is increasingly false for frontier models that ask clarifying questions, offer alternatives, and probe for missing intent. The advice still helps, but the underlying model of AI behavior is dated.
Read next: Recent work on clarifying-question behaviors in agentic models; Anthropic's writing on Claude's interrogative defaults; research on mixed-initiative interaction - incompleteThe four pre-prompt questions (purpose, audience, tone, success criteria) are classical rhetorical concerns and work well for human-output tasks like letters and memos. They do not cover technical prompting concerns specific to LLMs: context selection, decomposition of multi-step tasks, tool use, output format constraints, and verification strategies — all of which can matter more than tone or audience.
Read next: Prompt engineering guides that distinguish between generation tasks and agentic tasks; Anthropic and OpenAI documentation on context engineering; literature on chain-of-thought and decompositionclosed byPrompting Claude Fable 5 — Zinsser's four rhetorical pre-prompt questions omit the LLM-specific technical concerns (context selection, multi-step decomposition, tool use, output-format constraints, verification). This guide supplies exactly those: effort/context control, subagent decomposition, tool-use elicitation, readability/output-format instructions, and self-verification against tool results — matching the gap's reading_direction toward prompt-engineering guides that address agentic, not just generation, tasks. - incompleteSvenson excludes the exploratory use case in which the user genuinely does not yet know what they want and is using the AI as a thinking partner to find out. This is one of the most valuable uses of LLMs — and it is precisely the case where the Zinsser test would tell you not to prompt yet, which would be bad advice.
Read next: Andy Matuschak on tools for thought; literature on writing-to-think (Zinsser himself elsewhere acknowledges this); research on using LLMs as brainstorming partners - incompleteThe universal claim that intent-clarity dominates all technique is asserted but not measured. Different users at different skill levels likely face different bottlenecks — some are blocked by vague intent, others by context-loading, others by inability to evaluate outputs. The essay's one-size-fits-all framing under-serves users whose bottleneck is elsewhere.
Read next: Empirical studies on prompt-quality factors; segmentation of AI-user skill levels (novice/intermediate/advanced) in productivity research - incompleteThe analogy presumes that writing-craft principles transfer cleanly to prompting because both are 'communication.' But writing is human-to-human communication with shared context and reading conventions; prompting is human-to-model communication with very different priors, failure modes, and feedback loops. The essay does not stress-test where the analogy breaks.
Read next: Cognitive science of human-AI communication; research on shared common ground in conversation (Herbert Clark); the literature on prompting as distinct from natural-language instruction