5.8 KiB
artifact, name, layer, client, audit_id, date, prompt_library_version, engine_set_version, competitor_set_version, status, qa
| artifact | name | layer | client | audit_id | date | prompt_library_version | engine_set_version | competitor_set_version | status | qa | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| A | Diagnostic Record | 1-2 | draft | qa | approved |
|
Artifact A — Diagnostic Record
Purpose: Pure observation. Zero recommendations. This document must never reach a client raw — it is internal evidentiary backing behind the synthesized client report.
Scope, corrected after cross-check against path-to-poc-sequencing.md and existing repo templates: this file holds only the AI-visibility measurement content that has no existing home in the repo — the Core Visibility Index, Directional Context Indicators, Prompt-Level Observation Ledger, and per-engine sourcing context. It does not duplicate data-inventory-v1.md (Layer 1) or threat-register-v1.md (Layer 2). Findings from this file feed entries into those two documents using their existing schemas — this file is a source input to them, not a parallel record.
Banned-language gate (authored fields only — see field scope below): should, recommend(s)/recommendation(s), fix/fixing, opportunity/opportunities, package(s), proposal/propose, solution(s), action plan, next step. Automated token matching is a first-pass control; human review remains required for synonym evasion.
Gate field scope:
| Field type | Gate applies? | Reason |
|---|---|---|
| Executive summary, observation narrative, competitor observation summaries | Yes | Authored interpretation |
| Raw prompt text, raw engine output, source titles/URLs, platform field labels, quoted business names | No | Observed evidence or system metadata |
1. Scope and Method
- Client:
- Audit ID:
- Date:
- Prompt library version:
- Engine set version:
- Competitor set version:
- Prompt count:
- Engine count:
- Competitor set:
Method notes
Describe the sampling method, engine set, and limitations. (Not gated — this is methodology, not a finding.)
Limitations
"AI answer engines are dynamic. Results can vary based on model version, user session, location signals, personalization, and timing. This audit is a structured sample designed to identify visibility patterns. It is not a complete census of all AI responses and does not guarantee rankings, traffic, or revenue."
2. Run Inventory
| Run ID | Prompt ID | Engine | Timestamp | Raw output stored | Screenshot stored |
|---|---|---|---|---|---|
3. Core Visibility Index (quantitative, 70 pts)
Directly countable from captured audit runs only. No qualitative bands folded in.
| Component | Weight | Score |
|---|---|---|
| Mention rate | 35 | |
| Mention strength / position | 20 | |
| Description accuracy | 15 | |
| Total | 70 |
4. Directional Context Indicators (Tier 2, reported separately — not part of the 70)
| Indicator | Purpose | Evidence tier |
|---|---|---|
| Source/citation quality | Whether supporting sources favor client, competitor, or neither | Tier 2 |
| Competitor observation share | Relative appearance frequency across tested prompts | Tier 2 |
| Engine variance | Inconsistency across engines (see per-engine sourcing note, §6) | Tier 2 |
Client-facing language: "The Core Visibility Index is based on directly countable audit observations. The directional context indicators are qualitative and indicative, not a measured market share."
5. Prompt-Level Observation Ledger
| Prompt ID | Prompt text | Engine | Timestamp | Client observed | Mention strength | Accuracy state | Competitors observed | Source observations | Evidence tier |
|---|---|---|---|---|---|---|---|---|---|
6. Competitor Observation Summary
| Competitor | Observed count | Strong mentions | Moderate mentions | Weak mentions | Source observations | Notes |
|---|---|---|---|---|---|---|
| Per-engine sourcing context (Tier 2, indicative — source: Ty business-readiness review, 2026-08-06; requires direct audit observation or authoritative documentation to upgrade): |
| Engine | Primary local-answer source, working model | Highest-leverage signal, working model | Evidence tier |
|---|---|---|---|
| Gemini | Google's local graph | GBP, Maps/Places, reviews, categories, hours | Tier 2 |
| ChatGPT | Bing web index + partner place data + open web | Foursquare places data, Bing-visible consistency, business website, directory NAP | Tier 2 |
| Claude | Tools/APIs + web, often Google Places when looking up locals | Google Places/Maps-aligned data, clear website, consistent public facts | Tier 2 |
Do not score identical prompts across engines as if testing the same underlying signal — they source differently.
7. Feed-Forward
This file's findings do not get restated in a duplicate entity/evidence format. They feed forward directly into:
docs/clients/<client>/data-inventory-v1.md(Layer 1) — entity and signal state (GBP, citations, website, reviews) uses that file's existing schema, not a second one here.docs/clients/<client>/threat-register-v1.md(Layer 2) — evidence items with Tier 1/2 and Critical/Major/Minor severity are entered there directly, using that file's existing structure, not duplicated in this file.
Action for whoever processes an audit run: after completing §1–6 above, transcribe entity/signal observations into data-inventory-v1.md and evidence findings into threat-register-v1.md before this file is marked qa or approved. This file's own status should not be considered complete until that transcription is done.
8. QA Reference
- Evidence tier gate:
- Banned language gate (authored fields only):
- Human review gate:
- Reviewer:
- Review date:
- Transcribed into data-inventory-v1.md:
- Transcribed into threat-register-v1.md:
- Linked to Layer 1b owner-access record (if applicable):