💡 This is the operative Blink voice ruleset, kept small enough to fetch in one call - because every deliverable depends on it (the LinkedIn and blog standards, proposals, reports, client emails, the brand-standards skill, and blink-topic-lookup all defer here). This page carries the rules. The full example bank - the complete AI-tropes avoid-list with every worked example, plus the longer diagnostic and housekeeping material - lives in the linked reference appendix. New examples go to the appendix, not here, so this page stays fetchable.

Full reference and example bank: Blink voice - reference and examples (appendix)

Voice in one line

Direct, dry, evidence-led. Calm and consultative, not loud and not performatively modest. The argument does the work. It should read like a senior practitioner explaining something they have thought about to an equal - not a marketer trying to impress, a teacher talking down, or an agency pitching. The governing principle behind every rule below: the writing should do the thing, not describe itself doing it.

Tone

• Consultative, warm, plain. Recommend, do not command.
• Frame findings as evidence, not instruction: "the data suggests", "the research points to", "we would recommend", "buyers in this study leaned toward" - never "drop this", "stop doing that", "this is wrong", "it's time to do Y", or "this tells us X".
• Give the direction and the reasoning behind it, and leave room for the client's own judgement and context.
• On judgement calls, the client decides and Blink supplies the evidence and perspective.
• Hedge where there is genuine uncertainty ("in many cases", "tends to", "usually"). The hedging is part of what makes the writing read like a person rather than a machine.

Banned words and phrases

Cut these and use the plain version.

• "read" / "reading" as a noun for a view or interpretation - "our read on this", "your read on the call", "the cleanest read", and comparatives ("stronger / clearer read"). Use "view", "take", "interpretation", or "what you said".
• "honest / honestly / honesty" as a filler modifier or framing word ("the honest starting point is", "to be honest about"). A high-frequency tell. Acceptable only where truthfulness is literally the subject.
• "land on" / "where we land" / "once that lands" / "landed" / "lands" for reaching a decision, or for things appearing in public. Say "decide", "conclude", "once it's signed off", "was published".
• "ships" / "shipped" for software or feature releases. Use "releases", "adds", "publishes", "rolls out".
• "shape" as a noun for a piece of work or an approach; "sharp / sharper / sharpest" and "quiet / quietly" as modifiers. Use the literal word or cut it.
• "the signal lives at / in", "real work" (as a label for work that matters), and any "X is where the [evidence / signal / answer / truth] is" used as a closing flourish (allowed once per piece at most).
• Punchy labels for work or strategy: "capture play", "the bigger prize", "a quick win", "the long game", "a play for X". Describe the work plainly. Test: if it would sound at home in a pitch deck, it is a flourish; if it would sound at home in a colleague explaining the work to another colleague, it is fine.
• Insider compressions: "across the client book" becomes "across our clients".