brand-lint

name	brand-lint
description	DBA prevalence checker (Romaniuk ≥80%), banned-phrase + AI-slop + E-E-A-T scanner, GEO citation density check (≥1 citation per dollar amount, ≥1 statistic per H2). Runs as `pnpm brand:lint` for content data files. Used by Stage 5 Brand+Guardrail (binary pass/fail).
metadata	{"family":"ops","owner":"brand + content","last_reviewed":"2026-04-29T00:00:00.000Z","version":"1.0.0","related_skills":["marketing-foundations","content-review","ai-seo"],"kpis":["DBA prevalence ≥80% per asset (corpus-wide rolling scan)","100% of dollar amounts have inline citation","100% of H2 sections have ≥1 statistic","0 banned-phrase / AI-slop instances on shipped content"],"marketing_pillar":3,"seo_standard":"B","kpi_tier":3,"funnel_stage":"all","content_class":"all","maturity_stage":"descriptive","used_by_stages":[5]}

Brand Lint

Per-page + corpus-wide brand asset enforcement. Runs as pnpm brand:lint / pnpm brand:lint:all.

When to Use

• Stage 5 Brand+Guardrail per-content-commit (binary pass/fail).
• Weekly corpus scan (DBA prevalence rolling check).
• Pre-merge gate via lefthook on PRs touching content data files.

Do Not Use When

• Need editorial scoring (the 6-dimension rubric) — use content-review.
• Need editorial acceptance scoring — use content-review.

Per-Page Checks (Stage 5 input)

For every modified content file, scan and pass/fail:

1. Banned phrases (BRAND.md)

• Reject any banned phrase from nextjs-app/docs/BRAND.md banned-phrase registry.
• Examples: "best in the industry", "industry-leading", "revolutionary", "amazing", "game-changing".

2. AI-slop patterns

• Generic uncertainty ("studies show", "many experts agree") without named source.
• Listicles with vacuous bullets.
• "It's important to note that..." filler.
• Sentences starting with "Whether you're a..." (overused AI opening).

3. Citation density (GEO Pillar 1 — Aggarwal et al. KDD 2024)

• Scan body text for dollar amounts + percentages.
• For each: require an inline (Source Year) citation OR adjacent DataSourceCitation component.
• Fail if dollar_amounts > citations_count.

4. Statistic density per H2 (GEO Pillar 2)

• Parse content into H2 sections.
• For each H2: require ≥1 numeric statistic with unit/context.
• Fail if any H2 has 0 statistics.

5. Quotation requirement (GEO Pillar 3)

• For data-rich templates (role-page, comparison-page, original-research-page): require ≥1 quotation (worker testimonial, expert quote).
• Fail if 0 quotations on those templates.

6. Mandatory components per template

• AuthorByline present (DBA V2)?
• DataSourceCitation present where claims warrant (DBA V3)?
• ContentFreshness / "Last updated" visible (DBA V5)?
• FAQ-with-schema present (DBA S2)?
• InternalLinkHub present (DBA S4)?
• Contextual CTA present (DBA S5)?

7. Linguistic DBA enforcement

• L1: prefer "shifts" over "assignments"/"gigs" in non-technical sections (≥80% threshold per page)
• L2: "the app" over "the platform" in worker-facing copy
• L3: "Flexers" for community-context references
• L4: pay range first, then average

8. YMYL-specific (Pillar 3 financial + compliance)

• Reviewed-by line present?
• ≥3 Tier 1-2 citations per article (BLS / DOL / IRS / state agency)?
• Methodology link/callout?

Corpus-Wide Checks (Weekly via weekly-vital-loop)

For each DBA in marketing-foundations/reference/dbas.md:

prevalence = pages_with_dba / total_eligible_pages

• Pass: prevalence ≥ 80% (Romaniuk threshold)
• Warn: prevalence ≥ 60% and trending down
• Fail (corpus-wide warning to brand team): prevalence < 60%

Per-DBA targets at marketing-foundations/reference/excellence-targets.md Pillar 3 table.

Output Contract

Per-page (Stage 5 input)

Pass/fail JSON + markdown summary:

file: <path>
status: pass | fail
blocking_issues:
  - type: missing_citation | banned_phrase | ai_slop | missing_dba | etc
    line: <number>
    text: <quoted snippet>
    rule: <rule reference>
    fix: <suggested fix>
warnings:
  - <similar shape, non-blocking>

If status: fail, Stage 5 posts Slack APPROVAL gate per PIPELINE.md.

Corpus-wide (weekly)

Markdown at nextjs-app/docs/reports/<date>/brand-lint-corpus.md:

## Brand-Lint Corpus Scan — <date>

### DBA Prevalence
| DBA | Eligible pages | Pages with DBA | Prevalence | Status |
|---|---|---|---|---|
| V1 Flex Pink | 1560 | 1480 | 95% | PASS |
| V2 AuthorByline | 870 | 845 | 97% | PASS |
| V3 DataSourceCitation | 920 | 880 | 96% | PASS |
| ... | ... | ... | ... | ... |

### Trends (4-week rolling)
- <DBA>: trending up/down/flat

### Action Items
- (none)  / surface as Stage 6 template-update opportunity for next Discovery

Implementation

Runs as pnpm brand:lint. Intended for pre-push use on any change to:

• nextjs-app/src/features/*/data/**/*.ts
• nextjs-app/src/lib/data/articles/**/*.ts
• nextjs-app/src/lib/data/cities/**/*.ts
• Other content data files

Pre-push: per-page check on changed files. Weekly cron-equivalent via weekly-vital-loop: corpus-wide scan.

References

• marketing-foundations/reference/dbas.md
• marketing-foundations/reference/excellence-targets.md
• seo-foundations/reference/geo-methodology.md
• seo-foundations/reference/eeat-checklist.md
• nextjs-app/docs/BRAND.md