What Can and Cannot Be Measured in AI Search

Method status: These boundaries govern the NRLC measurement convention; they do not claim access to private engine telemetry.

Directly observable

• A URL or domain visibly cited in a saved answer
• An entity named without a visible citation
• The engine, interface, prompt, run date, and returned source list

Derived from a declared observation log

• Under the NRLC measurement convention, citation rate = C / P. C is the number of eligible prompts with a qualifying citation; P is the total number of eligible prompts evaluated. Every reported rate must identify the engine, prompt set, denominator, and observation window.
• NRLC defines mention rate as the share of eligible prompts where the monitored entity is named in the answer, whether or not a URL is cited. A mention without source attribution is not counted as a citation.
• Changes between equivalent observation windows with the same scope and eligibility rules

Not established by GSC or one answer

• ChatGPT, Claude, or Perplexity mention share
• Causal revenue or “AI traffic” from an answer impression
• A universal citation rate across engines with different prompt sets
• The private retrieval path or reasoning used to produce an answer

Methodology for defensible claims

State whether a result is directly observed, calculated from the observation log, inferred, or unavailable. Retain C and P with every citation rate, and disclose the engine, interface, prompt-set version, eligibility rules, and observation window.

Limitations

This method measures observed outputs, not private model reasoning. Results can vary by engine, interface, prompt wording, account state, and time. A citation occurrence does not prove that a source caused an answer, produced a click, or generated revenue. NRLC has not published a worked citation-rate example because there is not yet an observation log with a prompt count, engine, dates, citation count, or citation rate.