Dictating a Prompt with a Role and Audience

A spoken method for naming a role, describing a reader, and testing the brief.

Voiskey Voiskey Team
Voice PromptingChatGPT WorkflowDictating Prompt Role

Name the Role, Then Name the Reader

Dictating a Prompt with a Role and Audience starts with two sentences that most people skip. The first names the role the assistant should take and the responsibilities that come with it: a policy reviewer who checks claims against internal rules, a patient tutor who explains one idea at a time, or a skeptical editor who hunts for unsupported assertions. The second names the reader: what they already know, what decision they face, and how much time they will give the response. Say both before any task detail. Without a role, the assistant guesses a default voice; without an audience, it guesses a reading level.

Make the role functional rather than flattering. Instead of asking for a world-class expert, describe the work the role performs and the standard it applies. A useful formula is: act as this role, and you are responsible for this specific judgment, using this specific standard. Add what the role must refuse to do, such as inventing a citation or recommending a tool outside the approved list. Then say how the role should behave when information is missing: ask one clarifying question, state an assumption, or stop. Functional roles produce consistent output because the assistant has a job description. Flattering roles produce confident prose with no verifiable standard behind it, and the result is hard to review.

Describe the audience in terms of prior knowledge and intended action, not demographics. Say which terms they use daily and which need a definition, what they will do with the answer, and what would make the answer useless. For an executive, the action may be a go or no-go decision in one page. For a new teammate, the action may be completing a task without asking for help. Name the one question the reader will ask first, and ask the response to answer it early. A short glossary prevents quiet confusion. These details change structure more than adjectives do, because they tell the assistant what to put first.

Give one short sample of writing the audience already trusts. A paragraph from an internal memo or a help article is enough to show sentence length, level of detail, and how jargon is handled. Ask the response to match the sample register without copying its content. Then define what a good answer must contain: a recommendation, two supporting reasons, one risk, and a next step. Write these acceptance items before you dictate so you can check the response quickly. If the response satisfies the items but still feels wrong, the audience description is probably too thin; add one sentence about what the reader already believes.

Inventory What the Reader Already Knows

Split the audience description into three lists: terms the reader knows, terms that need a short definition, and terms that should never appear. Speak the lists rather than typing them, because naming a term aloud often reveals that it means different things to different people. For each defined term, give the reader version, not the textbook version. For each forbidden term, give the plain replacement. Attach this inventory to the prompt so the assistant does not reintroduce the jargon you excluded. A reader who knows the domain still benefits from a definition when your usage differs from the common one. Keep the inventory to one page so it stays usable during a live request.

Give the assistant one example the audience would accept and one it would reject, with a sentence explaining why. Concrete pairs teach register faster than a list of adjectives such as clear or professional. For a rejection, name the defect: too much background before the point, an unexplained acronym, a claim with no owner. Then ask the response to avoid that defect explicitly. Keep examples short so they act as calibration rather than material to copy. Review the pair with a colleague who knows the audience; if they disagree about the rejection, your audience description is not settled yet. Store the pair with the audience block so the calibration survives the next rewrite.

Say how the reader will consume the answer: on a phone between meetings, in a printed memo, or pasted into a ticket. That single fact determines length, heading count, and whether a table helps. Ask for a one-line summary at the top, then the shortest structure that supports the decision. If the reader will forward the answer, ask for a version that survives without your commentary. If the reader will discuss it live, ask for talking points rather than paragraphs. Format follows consumption, and consumption is part of the audience brief, not an afterthought. If both formats are needed, ask for the short version first and expand only on request.

Sometimes the role and the audience pull in different directions. A rigorous analyst writing for a hurried executive must choose what to leave out without lowering the standard. Handle the conflict in the prompt: say which value wins when they collide, and how the other value is preserved. For example, put the caveats in one short section instead of scattering them, or put the detailed reasoning in an appendix separated from the recommendation. Tell the assistant to flag any place where the compromise weakens the conclusion. Naming the tension prevents the response from averaging two incompatible briefs into a bland middle.

Speak the Constraints, Then the Style

Dictate constraints in the order that a reviewer would check them. Start with non-negotiables: length, required sections, facts that must be cited, and actions the reader must be able to take. Then add preferences: sentence length, examples, and level of detail. Style comes last, after the assistant knows what the response must accomplish. When style arrives first, it tends to dominate the result. Speak each constraint as a testable statement rather than a wish, and ask the assistant to list the constraints it understood before writing. Correct that list before accepting any draft. If two constraints conflict, ask which one the reviewer would relax first.

Turn the role into observable behavior. Instead of asking for a helpful assistant, ask for three behaviors: state the recommendation before the reasoning, define each term at first use, and separate facts from judgments. Behaviors can be checked; adjectives cannot. Add what the role does with uncertainty: name the missing information, give the best provisional answer, and state what would change it. This keeps the response useful without pretending to certainty. Test the behavior rules on a question with a known gap and confirm that the response names the gap rather than papering over it. Then remove any behavior you cannot observe in the output.

Set length in units the reader cares about: one screen, three short sections, or five bullets plus a recommendation. Vague limits such as brief are interpreted differently on every run. Pair the length with a shape: what comes first, what comes last, and what may be omitted. Ask for the most important sentence to appear in the first two lines. If the audience reads on a small screen, cap each section at a few sentences and avoid deeply nested lists. When you need a longer document, ask for an outline first, approve it, then request each section separately. Ask for the shape in the same sentence as the length so the response does not pad the middle.

List the moves the response must not make: no invented examples presented as real, no statistics without a source, no flattery at the start, no restating the question, and no closing paragraph that adds nothing. Prohibitions are especially useful when you dictate because speech tends to produce warm-up sentences. Ask the assistant to skip the warm-up and begin with the substance. Also prohibit changes to the audience brief: the reader does not become technical halfway through. When a draft breaks a prohibition, quote the sentence and ask for a revision that removes it, rather than asking for a general cleanup. Keep the list to five items; a long list of prohibitions dilutes the ones that matter.

Test the Brief on Two Readers

Voiskey lets you speak the role and audience block directly into the field where you are drafting, so the brief and the prompt stay together. Press the shortcut, say the role, the reader, the constraints, and the ask in one pass, and the text lands at the insertion point with filler and false starts removed. Then edit the two sentences that carry the most weight: the role responsibility and the reader action. Reading them aloud again is a fast check for ambiguity. Because the workflow is speak, polish, type, and edit, the brief is easy to revise before you send it. If a sentence sounds like a slogan rather than an instruction, replace it with the action you want the assistant to take.

Run the same prompt twice, changing only the audience block: once for a decision maker who will read for one minute, once for a practitioner who will act on the details. Compare the two outputs side by side. The structure should change, and the facts should not. If the facts drift between versions, the task description is too loose. If the structure does not change, the audience block is too vague. Keep both outputs and the two audience blocks as a calibration pair. This test takes a few minutes and reveals more than a long discussion about tone. Write down which structural change you expected before you compare, so the test measures your brief rather than your mood.

Read each output aloud as the intended reader. Stop at the first sentence that requires re-reading, and mark whether the cause is a missing definition, a buried recommendation, or an unexplained leap. Fix the audience block rather than patching the output sentence by sentence. If you find yourself editing more than a few lines, the brief is wrong, not the draft. Record which change to the brief produced the improvement. Keep the marked sentence and the brief change in the same note, with a date, so the lesson is retrievable months later. Over several rounds this record becomes a reliable recipe for that audience, and new team members can use it without reconstructing your reasoning.

If one response must serve two very different readers, split the deliverable instead of compromising the brief. Ask for a one-page decision summary and a separate working note, each with its own audience block. Keep shared facts identical and let structure diverge. Tell the assistant which document is primary and which is supporting. This is usually faster than negotiating a middle register that satisfies nobody. Review both against the same fact list before sending. Name the primary reader at the top of each document so a reviewer knows which standard applies. A split deliverable with a clear primary reader is easier to approve than a single document trying to be two things at once.

Keep Role and Audience Blocks Reusable

Save each role and audience block as a named note: decision maker, new teammate, external partner, skeptical reviewer. Keep the blocks short enough to read in one breath, because a block you cannot speak comfortably is a block you will not use. Add one line about what the reader already knows and one about the action they will take. Review the blocks quarterly and delete the ones nobody used. A small set of tested blocks beats a long taxonomy, and the act of reading them aloud keeps the language concrete. Put the name at the top of the note so it is easy to find later.

When several people write for the same audience, agree on one block and one sample paragraph. The team plan offers shared vocabulary and workflows, which keeps role definitions and reader assumptions aligned across writers. Store the block beside the sample so a new writer can see the standard in action. Ask each writer to run the two-reader test once before publishing. Record disagreements about the audience in the block itself rather than in a separate discussion. A shared block is only shared if everyone can find it, so keep one canonical copy. The block then carries the team judgment, and reviews focus on substance instead of renegotiating tone.

Voiskey learns names, jargon, acronyms, and spelling preferences, which matters when the audience block includes product names, internal program names, or a house style. Say the names once in the block and use them consistently; the app keeps the spelling you prefer. For readers in another language, ask for the response in that language and a short summary in yours. Voiskey supports more than one hundred languages and translation, so the role and audience brief can stay in the language you think in while the deliverable changes. Check the spelling in the final text once, because a single wrong apostrophe undermines a careful brief.

Download Voiskey for Mac and Windows and dictate one real brief before you build a library. Choose a message you must send today, name the role and the reader out loud, and speak the constraints in order. Edit the result, then save the block that worked. Repeat for a second audience and compare. Within an hour you will have two tested blocks and a clear sense of which sentences carry the brief. Keep the blocks where you draft, not in a separate notes system, so the next message starts from a tested brief instead of a blank field. Then keep the first two blocks visible for a week and refine the wording as you use them.

FAQ s

Voiskey is an AI voice typing app that works at the system level: press a shortcut, speak, and the text appears at the insertion point in any app or text field. It removes filler words, stutters, false starts, and self-corrections, so the role sentence, the reader description, and the constraints arrive as one clean prompt.

Yes. Voiskey adapts punctuation, formatting, tone, and wording to what you are trying to do, so the same request can sound like a decision memo for one reader and a working note for another. You describe the reader in the spoken prompt, and the app polishes the phrasing before the text lands at the insertion point.

Voiskey learns names, jargon, acronyms, and spelling preferences, so product names, program names, and house terms stay consistent across prompts. Say them in the role and audience block once and reuse the block. The app then places the polished prompt at your insertion point with your preferred spellings.

Yes. Voiskey supports more than one hundred languages and translation. You can dictate the role and reader block in one language, ask for the response in another, and request a short translation back for review. The spoken brief and the final text need not share a language, which helps when you write for readers outside your own.

Say less. Get more
Say less. Get more
Say less. Get more
Say less. Get more
VoiskeyVoiskey

Your voice, amplified.

Download Voiskey for Mac and Windows and dictate a role and audience brief.

Use it across macOS, Windows, iOS, and Android

Use it across