What Is Prompt Visibility?

Prompt visibility is a brand's presence in AI answers measured at the level of a single prompt: for "best accounting software for freelancers," how often is the brand mentioned, cited, or recommended when that exact question is asked? It is the atomic unit of AI visibility measurement — the analogue of a keyword ranking, rebuilt for engines that generate answers instead of ranking links.

Why does prompt-level tracking beat keyword tracking?

Keyword rankings assume a stable results page that every user sees. Generated answers break both assumptions: the "result" is synthesized fresh each time, and it names brands rather than listing URLs. Prompt-level measurement matches how the surface actually works. It also captures intent granularity keywords blur — "accounting software" is one keyword, but "best accounting software for freelancers," "QuickBooks alternatives for freelancers," and "is FreshBooks worth it" are three different battles with different winners. Building that question inventory is the job of prompt research.

What does a prompt visibility record contain?

For each tracked prompt, per engine, per run, a measurement system records:

• Mentioned — is the brand named in the answer body?
• Cited — is the brand's domain in the sources?
• Position — first mention, list rank, or trailing reference
• Framing — recommended, neutral, or caveated
• Competitors present — who else the answer names

Aggregated over runs, this yields a per-prompt visibility score that can be trended weekly and compared across engines.

What are the sampling caveats?

Two sources of noise demand respect. First, answer volatility: LLMs decode probabilistically (temperature and nucleus sampling), so the same prompt yields different brand sets across runs — visibility is a distribution, not a fact. Second, context sensitivity: answers shift with user location, account history, and model version, so measurement should control for region and re-baseline after engine model swaps. The practical rule: never act on a single run, and treat week-over-week trends from repeated sampling as the real signal — the methodology behind tracking AI mentions properly.

Example

A brand's aggregate visibility looks flat at 22%. Prompt-level data shows it winning "X vs Y" prompts at 80% but absent from all "best X for enterprise" prompts — invisible in the highest-value segment. The flat average concealed the exact gap worth fixing.