A worked example

What a useful AI visibility audit looks like.

This is a synthetic, non-client example. It shows the evidence trail behind an AI visibility audit, from a fictional stored answer record to a careful next action. The prompts, records and figures are illustrative only.

Read the boundary first

Nothing on this page is a live answer, a company result, a customer result, a demand estimate or a guarantee. It is teaching material built to make the reporting method inspectable.

Get a free AI Search audit → Read the measurement method

The Schmitdy dolphin checks answers and sources in a harbor town

Evidence record

01. Observed answer evidence

An audit starts with a stored record, not a score. This fictional record shows the fields a reviewer needs before they can interpret a conclusion.

OBSERVED

Stored teaching record, not a live engine response

Prompt version: Illustrative buyer question 01
Engine and market: Illustrative engine, UK English
Observation status: Valid teaching record
Visible source field: Present: two illustrative source labels

Illustrative answer excerpt

For this invented buyer question, the sample brand is one fictional option. The stored wording lets a reviewer inspect the mention before any conclusion is made.

Illustrative visible source fields

Fictional source label A
Fictional source label B

Everything in this record is invented teaching material. It shows the review structure without presenting an engine result, source or business as real evidence.

Derived metrics

02. Derived figures stay attached to their inputs

A percentage is useful only when a buyer can see the numerator, denominator and boundary around it. The three figures below are deliberately fictional calculations.

Illustrative calculation

Illustrative visibility

5 of 12

Five fictional valid records name the sample brand. 5 ÷ 12 = 41.7%.

Illustrative calculation

Illustrative source presence

3 of 5

Three fictional brand-bearing records show at least one visible supporting link.

Illustrative calculation

Checkable-claim coverage

4 of 5

Four fictional brand-bearing records contain a statement that could be checked. One has no checkable statement.

Review ledger

03. Keep four types of statement separate

A sceptical buyer should be able to point to every sentence and see whether it is observed, calculated, recommended or limited.

Statement type	What this sample shows	What a reviewer can do next
Observed answer evidence	A record preserves the prompt, market, language, run status, fictional answer excerpt and fictional source labels.	Read the captured fields and distinguish a stored mention from a conclusion.
Derived metric	The illustrative visibility rate comes from five fictional records out of twelve.	Recalculate the percentage from the displayed counts.
Recommendation	Review the seven fictional records without a sample-brand mention before changing a page.	Decide which buyer task deserves deeper evidence or content work.
Limitation	This page contains no analytics, booking or revenue evidence.	Leave commercial impact unavailable until an approved source can support it.

Limitations

04. Limits are part of the result

A clean audit does not use a blank field to imply a zero. It names what the record can and cannot establish.

This practice example does not measure traffic, enquiries, bookings or revenue.
A visible source link is not proof that it caused a brand mention.
A change after an intervention is an observation, not causal proof.
A result from one engine, market or language cannot stand in for another.

Evidence standards

Move from a sample to an evidence-led audit.

A live audit begins by agreeing the buyer questions, markets, engines, evidence rules and observation window. It keeps unavailable sources visible rather than filling the gap with a zero.

Read the measurement method →Explore public proof formats

FAQ

Questions about this sample

Is this a real audit result?+

No. Every prompt, record and figure on this page is synthetic teaching material. It is here to show the reporting structure, not to claim a result.

Why show fictional numbers at all?+

The fictional numbers make the arithmetic inspectable without exposing a company, a report or a measurement source. They are labelled illustrative and cannot be used as a benchmark.

What would make a live audit credible?+

A live audit needs a frozen question set, named engines and markets, stored answer evidence, defined calculations, an explicit missing-data rule and a reviewable connection to any commercial outcome.

Next step

Move from a sample to an evidence-led audit.

A live audit begins by agreeing the buyer questions, markets, engines, evidence rules and observation window. It keeps unavailable sources visible rather than filling the gap with a zero.

Get a free AI Search audit →