Office of Small Failures / Case File 004 Priority test
AI search measurement experiment

Signal
Sample

Measure AI-search visibility without pretending AI has rankings. Build a fixed panel of customer questions, test them manually in AI answer systems, and turn the observations into repeatable evidence.

Traditional search gives us rankings. AI answer systems give us variable observations. Signal Sample makes those observations systematic enough to compare — without manufacturing certainty.

Set the sample

01
These become selectable competitor observations.

Customer question panel

02

Aim for roughly 10–20 stable questions that reflect real customer decisions. Keep the panel consistent between rounds if you want meaningful comparisons.

Record an observation

03
On desktop use Ctrl/Cmd to select multiple. On mobile, tap the competitors that appeared.

Evidence log

04
DateRoundPromptSurfaceBrandCitedCompetitorsAccuracyEvidence
Method boundary: Signal Sample does not query AI systems for you, estimate prompt volumes, infer rankings or create a composite “AI visibility score”. It reports only what was observed in the sample you defined.

How to measure AI search visibility

AI visibility tracking is difficult because ChatGPT, Gemini, Perplexity and other answer systems can return different answers across prompts, users, dates and regenerations. A single screenshot is evidence of one response, not a universal ranking.

Signal Sample is a lightweight AI search measurement framework: choose a stable panel of customer questions, manually test the same questions across the AI surfaces that matter to you, and record brand mentions, citations, competitor appearances and factual accuracy. Repeating the same panel creates a more defensible way to discuss change over time.

What does an AI visibility tracker measure?

This v0 measures presence rate, citation rate, competitor presence and accuracy problems inside your defined sample. It can also compare one measurement round with the previous round. These are observations, not estimates of total ChatGPT brand mentions or universal LLM visibility.

Why not create an AI visibility score?

A proprietary score can imply precision that the underlying observations do not support. Signal Sample keeps the components visible so stakeholders can see the sample size, method and evidence behind the result.

Also testing SEO priorities? Try WorthDoing →