Copy is clear — one idea per sentence, simple present tense, no double negatives, no noun stacks, short words, acronyms and technical terms have a reachable definition

Requirement

Copy must be clear enough to parse in one read. Six mechanics enforce this:

1. One idea per sentence. Keep each sentence focused on a single action or fact.
2. Simple present tense where possible. Prefer direct, present-tense constructions: "Application submitted" not "Application has been submitted."
3. No double negatives. Use the positive form when one exists: "I want to receive reminders" not "I do not want to opt out of reminders."
4. No noun stacks. Several nouns in a row make meaning harder to scan. Use verbs for actions, or adjectives to describe the main noun. For example, write "Review support cases" instead of "Support case review."
5. Short and simple words. Prefer the plain form when one exists:

6. Acronyms and technical terms have a reachable definition. Any acronym or unfamiliar technical term must let a user find out what it means without leaving the task. It need not be expanded on first use — an inline expansion ("Know Your Customer (KYC)"), a tooltip or info affordance, help text, or a glossary link all satisfy this. The test is that the explanation is discoverable, not where it sits. Established terms your users genuinely use every day need none.

This applies wherever copy appears as prose — instructional text, help copy, confirmations, empty states, and error messages. Labels and fragments are not sentences and are not in scope for clauses 1–3; clauses 4–6 apply to labels too.

See also: CNT-2 catches unexplained acronyms and jargon in names — navigation items and page titles, at intent/plan time. Clause 6 here is the body-prose sibling: an acronym or technical term inside running copy whose definition a user cannot reach. A surface can pass CNT-2 (every name is plain) and still fail clause 6 (a body-copy acronym with no tooltip, help text, or glossary behind it).

Rationale

Clarity is the terminal quality goal for UI copy. These six mechanics each reduce parse time. They are bundled under one control because they share a single test: can a user understand this in one read without backtracking? A waiver on one clause should not silently cover a failure in another, so findings must name the clause. An acronym a user cannot decode stops the read as surely as a two-idea sentence, so the reachable-definition rule belongs in the same bundle.

Bundled (not split) by analogy with CNT-3, which bundles second person, active voice, and the 25-word ceiling under one ID. The word-list clause is the only one with a deterministic half; the rest are judgment.

Passes when

• "Select a project to begin." — one idea, present tense, active, no stacking.
• "Application submitted." — simple present, concise.
• "I want to receive reminders." — no double negative.
• "Review support cases." — verb for action, not a noun stack.
• "Try again." — plain word, not "Attempt to retry."
• "Know Your Customer (KYC)", or "KYC" beside an info affordance that expands to the full term — the definition is reachable.

Fails when

• A sentence packs two instructions: "Select a project and then choose a date to begin."
• Past-perfect where present fits: "The form has been submitted."
• Double negative: "Do not opt out of reminders."
• Noun stack: "Customer support case review status."
• Long word with a plain substitute: "Please attempt to initiate the process."
• An acronym or technical term in body copy with no reachable definition: "Complete the KYC review" where nothing on the surface — no expansion, tooltip, help text, or glossary — tells a user what KYC means.

How to verify

Deterministic half — checks/content-lint.py (planned extension): scan copy strings for words in the substitution table and flag each match. Judgment half — the evaluator reads each sentence and checks clauses 1–5 in order, then checks clause 6 against the rendered surface (an affordance such as a tooltip counts, so the copy string alone is not enough to judge it), as described below.

Evaluator guidance

Quote the offending phrase in every finding. Name the clause (1–6) that was violated. Give the rewrite.

Flag:

• Clause 1 — a sentence with two distinct ideas joined by "and" or a semicolon, where splitting would not change the meaning.
• Clause 2 — past-perfect or passive tense where a simple present alternative exists ("has been submitted" → "submitted"; "will be notified" → "you'll be notified").
• Clause 3 — a negated verb paired with a negated noun or negative prefix ("do not opt out", "cannot fail to", "not uncommon").
• Clause 4 — three or more nouns in a row where the first or middle noun could become a verb or adjective ("attendance report submission" → "submit the attendance report").
• Clause 5 — a word from the substitution table when the plain form fits the context.
• Clause 6 — an acronym or unfamiliar technical term in body copy with no reachable definition anywhere on the surface (no inline expansion, tooltip, help text, or glossary).

Do not flag:

• Settled product names, programme names, or quoted policy text carrying a dx-waive CNT-9 reason="..." — confirm the waiver and move on.
• Labels and fragments for clauses 1–3 — "Save", "Due date", "Task Planner" are not sentences.
• Technical terms with no plain equivalent — "utilise" flagged only when "use" works; if the technical form is the precise term, it stands.
• Established terms your users genuinely use every day, which need no expansion (clause 6), and any term whose definition is one tooltip, help link, or glossary entry away.
• Deliberate past tense for tone: "Your session has ended" (closure framing) is acceptable when the alternative sounds abrupt. Document with a waiver.

Waiver

rationale (L2) — inline dx-waive CNT-9 reason="..." at the deviation site. Name the clause being waived in the reason. Policy-mandated verbatim text is the canonical case.