brag-evidence-writer

name: brag-evidence-writer description: > Rewrites or generates web content using the AiVIS Cite Ledger BRAG protocol, Based-Retrieval-Auditable-Grading. Transforms vague, promotional, or AI-invisible content into concise, factual, extraction-ready copy that AI answer engines can parse, verify, and cite. Produces direct answer blocks, entity definition paragraphs, FAQ sections, and methodology documentation. Use this skill whenever the user wants to rewrite page content for AI readability, improve their citation score, make content "AI-friendly", write copy that ChatGPT or Perplexity will quote, or when an AEO audit returns low Content Depth or AI Readability scores. Trigger on "rewrite for AI", "make this citable", "AI-readable content", "answer engine optimization", or when content is described as vague, promotional, or thin.

BRAG Evidence Writer

Transforms content into BRAG-compliant, AI-citable copy using the AiVIS Cite Ledger evidence protocol.

What BRAG means in practice

Every content unit must answer three questions simultaneously:

1. What exactly is wrong / what exactly is this, specific, not general
2. Where is the evidence, observable, verifiable from the page
3. What specific action/fact, implementable or quotable without context

Content that fails any of these three is not BRAG-compliant and will be deprioritized or skipped by AI answer engines during extraction.

Content transformation rules

Rule 1: Replace promotional phrasing with factual statements

Rule 2: Every claim must be specific and verifiable

Bad: "Schema markup is important for AI visibility." Good: "Pages with complete JSON-LD schema (Organization + WebPage + Speakable) score an average of 22 points higher on the AiVIS Cite Ledger schema dimension than pages with no structured data."

Bad: "We've helped hundreds of companies improve AI visibility." Good: "AiVIS Cite Ledger audits measure citation readiness across ChatGPT, Perplexity, Gemini, and Google AI Overviews using ten independently scored families (avs-v3)."

Rule 3: Direct answer blocks, Q then A within two sentences

This is the highest-value AI readability signal. Structure:

### What does [product] measure?
[Product] measures [specific thing] by [specific method]. [One sentence of supporting detail with a specific fact or number.]

The answer must be self-contained, an AI engine must be able to quote it without needing surrounding context.

Rule 4: Entity definition paragraph (canonical block)

Every page's first substantive paragraph must be a complete entity definition:

• Who/what the brand or product is
• What it does (verb + object, specific)
• Who it's for (audience)
• What makes it distinct (differentiator)

Template:

[Brand] is a [category] that [specific action verb + object] for [audience].
Unlike [category default], [Brand] [specific differentiator with evidence].
[One concrete fact that is verifiable: number, method, named output].

Rule 5: Methodology and trust content

AI answer engines weight pages more heavily that include:

• A methodology or how-it-works section with named steps
• Links to privacy, terms, and about pages (trust verification)
• Author or organization attribution
• Dates (published, modified, version)

When writing methodology content, always name the steps and give each a label. Unnamed steps ("first we do X, then Y") score lower than labeled steps ("Phase 1: Extraction, Phase 2: Alignment, Phase 3: Validation").

Content types and templates

1. Homepage hero (entity definition + canonical block)

# [Specific, non-generic H1 that includes brand name + primary function]

[Brand] is a [specific category] that [verb + what it does] for [audience].
[Specific differentiator]. [Named output or metric, what the user gets].

[Second paragraph: how it works in one to two sentences with a named method or framework.]

2. Feature section (extraction-ready)

## [Feature name, specific, not "Powerful Analytics"]

[Feature name] [does specific thing] by [named method].
[One concrete output: what the user receives, in measurable terms].

[Optional: one sentence on who this is for and when to use it.]

3. FAQ block (highest citability format)

### [Exact question a user would type, include primary term]

[Direct answer in one to two sentences. Self-contained. No "as mentioned above" or "it depends".]
[Optional: one sentence of supporting evidence or context.]

4. Methodology / how it works

## How [brand/feature] works

[Brand/feature] operates in [N] stages:

**Stage 1: [Named stage],** [What happens, specifically. Input → output.]
**Stage 2: [Named stage],** [What happens. What the system checks for.]
**Stage 3: [Named stage],** [What is produced. How it is validated.]

[One paragraph summarizing the full cycle and why this sequence matters.]

5. Scoring or grading content (high citability)

Tables, tier lists, and named scoring systems are among the most-cited content formats in AI answers. Always include:

• Named tiers with specific thresholds
• Observable criteria for each tier
• Attribution to a named framework or system

Quality gate before outputting

Run content through these checks:

• No sentence starts with "Our" followed by an adjective
• Every claim has a specific fact, number, or named element
• Every FAQ answer is self-contained (reads correctly in isolation)
• Entity definition paragraph present on page-level content
• No sentence requires surrounding context to be meaningful
• Methodology steps are named, not just described

Output format

For rewrites: show before/after in a comparison block, then explain the BRAG gate each rewrite satisfies.

For new content: output ready-to-publish markdown with a brief note on which AI readability signals each section targets.