Sample AI visibility audit report (sanitized / mock)

v1.2 · 19 June 2026 · Last updated 2026-06-19. This is a sanitized, fictional Snapshot audit. Numbers, citations, brand mentions, and GEO scores are mock. It is not a real customer report and it is not a measurement promise for any real brand.

Key takeaways

This page shows the shape of a Clarivy Snapshot deliverable, not a live customer's result.
The sample has 35 mock datapoints: 5 illustrative queries x 7 production endpoint labels.
The sample shows two deterministic GEO score fields: AI Citation Readiness and Source Trust Baseline.
A paid Snapshot uses the same file structure but contains your real brand, real queries, real responses, and a real audit-log URL.

MOCK sample. Read this before you draw any conclusion from the numbers.

The subject brand is Example Cross-Border Brand (URL placeholder: example.com), a fictional name reserved for documentation per IANA reserved domains. It has no real AI presence and no real audit data.
The 5 user queries in this Snapshot are illustrative; they are not from a real customer briefing.
The 35 datapoints (5 queries × 7 endpoints) are generated by the audit-runner mock orchestrator (RUN_MODE=mock). Real customer runs use the same template, the same Data Provenance block, the same 6 anti-hallucination rules — but real responses from the real LLM APIs.
All URLs (raw JSON, audit logs) point to example.test/2026-06-15/ord_sample_*/. They do not resolve — they exist only to show the deliverable structure.
Every link, button, and CTA in the sample leads to a mailto: address or the public Methodology page, not to a real order or real customer data.

What this sample shows — and what it does not

The goal of this public sample is to show the delivery structure of a paid Clarivy audit, end to end, without exposing any real customer:

It shows the 3-file delivery package (human PDF, AI-readable Markdown brief, raw-data index) and how a buyer (or the buyer's AI agent / Obsidian vault) consumes each file.
It shows the YAML frontmatter, stable finding_id / evidence_id / action_id identifiers, and the 13-section AI-readable Markdown brief structure.
It shows the deterministic GEO score pair used by Clarivy reports: AI Citation Readiness and Source Trust Baseline. These are derived from observed LLM responses and citation fields, not from SEO, backlink, traffic, CORE-EEAT, or CITE scoring.
It shows the Data Provenance footer (runner, endpoint list, sampled_at, inputs, outputs) and the stochasticity disclaimer.
It does not show what a paid audit's actual mention counts, citation map, or priority actions will look like for your brand — those depend on (a) which 5 queries you specify, (b) which endpoints return a result on the day of the run, and (c) the per-datapoint LLM response on that day.
It does not predict future rankings, lift, or business outcomes.

How to read this sample

Start with sample-human-report.html to see the leadership summary.
Open sample-ai-employee-brief.md to see how findings, evidence, and actions are cross-referenced.
Check sample-raw-data-index.md to understand how every datapoint is listed for verification.

What is mock vs what is real

Element	Status in this sample	Status in a paid Snapshot
Brand	Fictional placeholder	Your real brand URL
Queries	Illustrative prompts	5 queries you specify
Responses	Mock output from the audit-runner	Real responses from the 7 production endpoints
Audit-log URLs	Placeholder paths that do not resolve	Real public audit-log URLs
GEO score fields	Mock values derived from the mock datapoints	Deterministic values derived from your observed responses and citation fields
File structure	Real delivery structure	Same structure with real run data

The 3 files in this sample (Snapshot product, 35 datapoints)

File	Audience	What it is
sample-human-report.html	Leadership (CMO, Head of Growth, dev lead)	Browsable HTML alternative to the 1-page human-report.pdf. Shows the Snapshot cover page, per-endpoint score table, top-3 priority actions, top-3 cited domains, and the 2 mandatory footers (Data Provenance + stochasticity).
sample-ai-employee-brief.md	Internal AI agents, AI assistants, Obsidian vault	13-section Markdown Brief with YAML frontmatter. Stable `finding_id` / `evidence_id` / `action_id` identifiers, Obsidian `[[id]]` crosslinks, deterministic GEO score fields, and an instructions block that tells the buyer's AI how to summarize without paraphrasing.
sample-raw-data-index.md	Auditor, buyer, anyone verifying the report	Markdown index of every (endpoint, prompt) datapoint with a direct URL to the raw JSON. In this mock sample the URLs are placeholders under `example.test`; in a paid run they resolve to the public agentgeek-geo/audit-logs repo.

What you should compare against a paid run

When you receive a paid Snapshot ($149, the only product available today) audit, the same three files arrive in the same shape as this sample — but with:

Your real brand URL (not example.com).
Your 5 real queries (for Snapshot) or the standard 120-prompt matrix (for Standard).
Real LLM responses from the 7 production endpoints, with the exact model id recorded in the Data Provenance block.
A real auditLogsUrl pointing to a dated folder in the public audit-logs repo where every raw JSON is clickable.
A deterministic confidence: high | medium | low in the AI-readable Markdown brief frontmatter (see methodology.html §11.2).

What you should NOT conclude from this sample

The mention counts in this sample are 0 / 35 because example.com is a fictional domain with no real AI presence. A real brand audit will produce different numbers; the 0/35 baseline is not a forecast for your brand.
The 3 priority actions in the sample are placeholders. A real audit's actions are derived from the per-(endpoint, prompt) findings of the actual run.
The Citation Source Map in the sample is empty (no real LLM cited anything for example.com). A real audit will show real cited domains, where the LLM actually got its answer from.

FAQ

Is this sample a real customer report?

No. The public sample is sanitized and fictional. The numbers, citations, and brand mentions are mock and do not represent a real customer.

What does the sample show?

The sample shows the 3-file delivery structure: a human report, an AI-readable Markdown brief, and a raw-data index.

How is a paid Snapshot different?

A paid Snapshot uses the buyer's real brand URL, 5 real queries, real responses from the 7 production endpoints, and a real audit-log URL.

Can AI assistants read the sample?

Yes. The AI-readable Markdown brief is designed to show stable finding, evidence, and action identifiers so an internal AI assistant can summarize without losing traceability.

How the public sample is kept honest

This page is tracked in the internal Trust Evidence Register. The register's §2.8 "Public sample" rows cover 4 claims, each with a verification method and a risk level. The test suite audit-runner/test/public-trust-pages.test.js asserts on every commit that:

This index.html + the 3 sample files all exist on disk.
This page contains an explicit "this is a mock / sanitized sample" disclaimer (the orange-bordered box at the top of this page).
The sample files do not contain any real customer name, real customer email, real R2 signed URL, or any reference to a real customer domain (the runtime test in audit-runner/test/public-trust-pages.test.js scans for any such pattern and fails CI on a hit).
The sample files do not contain any banned phrase (no "guaranteed accuracy", no "always newest", no "best model", no "most advanced model", no "all AI models", no "ZDR (default)" / "default ZDR").
The landing page has at least one link to /audit/sample/, so a first-time visitor can find it without searching.

/audit/methodology.html — the technical contract (product matrix, 3-file delivery, Snapshot vs Standard/Enterprise, API vs product surface, Model freshness policy, Delivery QA, the 6 anti-hallucination rules, §11 confidence/accuracy/limitations).
/ai-search-visibility-audit/ — the core service page for Snapshot.
/geo-audit/ — the GEO audit explainer.
/audit/faq.html — the buyer-objection FAQ (7 commercial questions, bilingual).
/audit/self-audit-01.html — the Day-1 self-audit (historical Day-1 dogfooding artifact, 30 prompts × 7-9 endpoint experimental matrix; current production matrix is different). Kept for transparency. Not a customer-style sample; this is an internal dogfooding record.
/#pricing — the Snapshot Audit ($149, available today) and the Coming-soon Standard / Enterprise placeholders.

Changelog

v1.0 (2026-06-15) — initial release. 3-file mock sample at /audit/sample/: sample-human-report.html, sample-ai-employee-brief.md, sample-raw-data-index.md. Fictional brand "Example Cross-Border Brand" (URL example.com). Mock Snapshot product (35 datapoints, 7 endpoints). All URLs placeholder under example.test. No real customer data, no real R2 signed URL, no reference to any real customer domain.
v1.1 (2026-06-18) — P0 SEO/GEO stopgap. Title and H1 updated to "Sample AI visibility audit report" so the page better answers AI-search-intent queries. Mock content and the 3-file structure are unchanged.