# OSHUN V1 Veritas Content Tone Standards

Status: Accepted v1.0  
Owner: OSHUN Design + Product + Engineering  
Last Updated: March 26, 2026  
Implements: `V1-UX-037`  
Related:

- `docs/releases/v1/design/ux-principles.md`
- `docs/releases/v1/design/readability-standards.md`
- `docs/releases/v1/scope/immersive-vs-utilitarian-posture-rules.md`
- `docs/releases/v1/scope/scan-first-content-hierarchy-rules.md`
- `docs/releases/v1/specs/web-app-admin-v1-prd.md`
- `docs/releases/v1/specs/product-brief.md`
- `apps/oshun/web/src/components/domains/VeritasSurface.tsx`
- `apps/oshun/web/src/components/domains/veritas/VeritasFactChecker.tsx`
- `apps/oshun/web/src/components/domains/veritas/VeritasArticleReader.tsx`
- `apps/oshun/web/src/components/domains/veritas/VeritasRAG.tsx`
- `apps/veritas/web/src/messages/en.json`
- `apps/veritas/mobile/src/i18n/locales/en.ts`
- `docs/releases/v1/specs/todos.md`

## 1. Purpose

This document defines the canonical content tone for Veritas grounded surfaces
in OSHUN V1.

Veritas is the product's epistemic clarity layer. Its job is not to sound
clever, dramatic, or neutral-for-performance. Its job is to help the user
understand what is known, what is uncertain, what evidence supports the claim,
and what source posture applies.

The Veritas tone standard exists so grounded surfaces feel:

- precise
- proportionate
- trustworthy
- readable
- transparent about uncertainty

without becoming:

- sensational
- smug
- evasive
- false-balanced
- overly academic

## 2. Current Repo Signals And Gap

Current implementation review already shows a strong Veritas voice pattern:

- `VeritasSurface.tsx` uses framing such as `Truth-first stories`,
  `Every claim annotated with confidence context, source provenance, and verification trails`,
  and `Verify claims with evidence`.
- `VeritasFactChecker.tsx` centers verdicts, confidence, evidence strength,
  corroboration, and source credibility instead of mood-heavy storytelling.
- `READABILITY_STANDARDS.md` explicitly requires the sequence
  `claim headline, confidence/evidence, source, action`.
- `OSHUN_V1_IMMERSIVE_VS_UTILITARIAN_POSTURE_RULES.md` makes Veritas utilitarian
  by default and forbids atmospheric drift that hides evidence.
- `OSHUN_WEB_APP_ADMIN_V1_PRD.md` defines Veritas as `epistemic clarity` and
  requires explanations with traceability and current events interpreted with
  care.
- The standalone Veritas apps already use language such as `Fact Check`,
  `Bias Transparency`, `Sources`, `Human Reviewed`, `Breaking News`, and
  `Catch up on today's top stories`, which is factual and direct.

The gap is canonical tone governance. The repo has good raw instincts, but not
yet one explicit standard for how certain Veritas is allowed to sound, how it
should handle partial knowledge, and what kinds of media-style rhetoric are
non-canonical.

## 3. Canonical Tone Thesis

Veritas should sound like a careful editor and evidence analyst working in the
open.

The canonical Veritas voice is:

- clear before clever
- measured before dramatic
- traceable before persuasive
- direct before ornamental
- readable before academic

Veritas is allowed to sound confident when the evidence is strong.

Veritas is not allowed to:

- sound more certain than the evidence
- imitate cable-news urgency for ordinary updates
- inflate ambiguity into suspense
- bury uncertainty in small print
- perform balance where the evidence is not balanced

## 4. Tone Pillars

### 4.1 Precision Over Vibe

Veritas copy should prioritize exact meaning.

If a stronger or flashier phrase adds emotion but reduces epistemic accuracy, it
is off-tone.

### 4.2 Confidence Must Match Evidence

Veritas may use firm language only when evidence quality, provenance, and review
status justify it.

When evidence is mixed, limited, pending, or contradictory, the language must
show that clearly.

### 4.3 Transparency Over Performance

Users should be able to tell:

- what claim is being evaluated
- where the evidence comes from
- how strong the evidence is
- what remains unresolved

Veritas must never hide uncertainty behind polished summary language.

### 4.4 Neutrality Means Proportion, Not Both-Sides Theater

Veritas should not use emotionally loaded copy or ideological shorthand.

But it also should not pretend every position deserves equal framing when the
evidence clearly does not. Proportionality matters more than cosmetic symmetry.

### 4.5 Readable, Not Academic Fog

Veritas can be serious without becoming dense.

Users should not need to decode institutional prose just to understand the
result of a claim review or a source profile.

## 5. Language Rules

### 5.1 Sentence Shape

- Headlines should be factual and compact.
- Summaries should usually be one clear sentence before secondary detail.
- Explanations should move from conclusion to evidence, not the other way
  around.
- Do not stack multiple hedges in the same sentence if one precise qualifier
  will do.

### 5.2 Preferred Verbs

Prefer verbs such as:

- shows
- indicates
- reports
- verified
- disputed
- corroborates
- contradicts
- documents
- cites
- traces
- compares

These verbs are concrete and evidence-friendly.

### 5.3 Qualifier Rules

Use qualifiers deliberately:

- `verified` only when the evidence and review posture justify it
- `likely` when the evidence leans clearly but is not final
- `inconclusive` when the record does not support a stronger claim
- `reported` when describing a source claim that has not yet been independently
  confirmed
- `under review` when verification work is still active

Avoid vague filler qualifiers such as:

- `somewhat`
- `arguably`
- `basically`
- `kind of`

### 5.4 Vocabulary To Avoid

Avoid language such as:

- `shocking`
- `you won't believe`
- `explosive truth`
- `slam dunk proof`
- `undeniable` unless it truly is
- `everyone is saying`
- `massive cover-up` without direct evidence
- `game-changing` as default story shorthand

Also avoid self-important AI language such as:

- `Veritas has definitively solved this`
- `our model knows the truth`

Veritas can explain analysis. It must not mystify it.

## 6. Surface-Specific Tone Rules

### 6.1 Headlines And Claim Titles

Goal:

- tell the reader what the item is about in the fewest accurate words

Rules:

- Prefer factual structure over provocative framing.
- Name the subject, action, or finding directly.
- Do not hide the central claim in metaphor or tease copy.
- If the status is uncertain, reflect that in the headline or nearby label.

Good:

- `Arctic ice coverage reaches new seasonal low in satellite record`
- `Claim under review: vaccine shortage figures vary across agencies`

Avoid:

- `A climate bombshell no one can ignore`

### 6.2 Story Summaries And Explainers

Goal:

- explain significance without overstating

Rules:

- Lead with the key takeaway.
- Follow quickly with source, confidence, or evidence posture.
- Avoid opinionated framing words unless the surface is explicitly analysis or
  commentary and is labeled as such.
- Summary language must not outrun the underlying reporting.

### 6.3 Verdicts, Confidence, And Method Labels

Goal:

- make epistemic posture legible

Rules:

- Verdict labels must be plain and stable.
- Confidence language must be consistent across claim views, article readers,
  and AI-grounded answer surfaces.
- Method copy should explain how the system arrived at the result without
  overselling the mechanism.
- If confidence is low or mixed, the copy should say what further evidence would
  strengthen the conclusion.

### 6.4 Source And Provenance Language

Goal:

- help the reader evaluate trust quickly

Rules:

- Source names should be explicit and easy to scan.
- Provenance labels should sound factual, not promotional.
- Bias, ownership, corrections, and transparency language should stay sober and
  specific.
- Do not use vague praise such as `top-tier source` when a concrete metric or
  profile label is available.

### 6.5 Breaking And Developing Updates

Goal:

- communicate urgency without sensational spillover

Rules:

- Use urgency only when the story is genuinely time-critical.
- Short update language is acceptable, but still needs factual discipline.
- If information is preliminary, say so immediately.
- Do not reuse breaking-news tone for ordinary updates, newsletters, or stale
  events.

### 6.6 AI-Grounded Answers And Assistive Research

Goal:

- keep the answer accountable to evidence

Rules:

- Lead with the answer or current best assessment.
- Follow with citations, confidence, or source context.
- Make model uncertainty explicit instead of smoothing it into confident prose.
- Avoid anthropomorphic self-praise or magical-comprehension wording.

Good:

- `Based on the currently cited sources, this claim remains inconclusive.`

Avoid:

- `I have uncovered the real truth behind this claim.`

### 6.7 Corrections, Updates, And Uncertainty

Goal:

- maintain trust by being explicit

Rules:

- Correction copy should be direct, timestamped, and free of defensiveness.
- When new evidence changes a previous posture, say what changed.
- When data is missing, unavailable, or disputed, say that plainly.
- Do not hide uncertainty in euphemisms like `nuanced` when the real issue is
  lack of confirmation.

### 6.8 Search, Empty, And Navigation Copy

Goal:

- orient without drama

Rules:

- Search and empty-state copy should stay plain, helpful, and low-friction.
- Suggest broader terms, related categories, or evidence filters instead of
  vague encouragement.
- Labels such as `Fact Check`, `Sources`, `Bias Transparency`, and `Coverage`
  are preferred because they are exact and predictable.

### 6.9 Notifications, Briefings, And Newsletters

Goal:

- deliver concise context, not adrenaline

Rules:

- Briefing copy should sound useful and digestible.
- Notification headlines should reserve all-caps and urgent labels for truly
  breaking cases.
- Newsletter and digest copy can feel curated, but not glossy or promotional.

Good:

- `Catch up on today's top stories.`
- `Breaking: vote count delayed in two districts, officials cite verification checks.`

## 7. Do And Do Not Examples

| Scenario         | Do                                                                         | Do not                                        |
| ---------------- | -------------------------------------------------------------------------- | --------------------------------------------- |
| story headline   | `Court releases audit summary on procurement review`                       | `Bombshell audit rocks the establishment`     |
| claim verdict    | `Inconclusive: available filings conflict on the total`                    | `Mystery deepens around the real number`      |
| evidence summary | `Three primary sources support the timeline; one detail remains disputed.` | `Everything points in one obvious direction.` |
| source label     | `Source: Nature Medicine`                                                  | `Elite science outlet`                        |
| AI answer        | `Current evidence supports part of the claim, but not the full statement.` | `The AI has determined the truth.`            |
| breaking update  | `Developing: officials confirm two injuries; cause not yet verified.`      | `Chaos erupts as shocking details emerge`     |
| correction       | `Updated at 14:10 to reflect the revised ministry total.`                  | `A minor clarification has been made`         |

## 8. Definition Of Done For Veritas Copy

Veritas copy meets the OSHUN V1 tone standard only when it is:

- precise enough to withstand scrutiny
- proportionate to the evidence
- explicit about confidence and uncertainty
- readable without jargon overload
- serious without melodrama
- neutral in posture without false-balance theater
- traceable enough that the reader can understand why the claim is framed that
  way

If Veritas copy sounds like clickbait, advocacy disguised as certainty, or a
research paper abstract with the nouns removed, it is off-tone.
