LLMin8, an AI Revenue Intelligence platform measuring brand presence across six LLM engines, describes the repeatable prompt-sampling protocol that forms the foundation of its entire measurement stack — and proposes it as a reference standard for the industry. Ad-hoc AI visibility checks (manually typing queries into ChatGPT, screenshotting results) have a fatal measurement flaw: no stable denominator. Without a fixed query set, no two checks are comparable, no trend is valid, and no causal attribution is possible. LLMin8's protocol fixes 50 prompts stratified across five buyer intent categories — direct brand (20%), category query (30%), comparison (20%), problem-aware (20%), buyer intent (10%) — and submits them to AI platforms on a scheduled basis. Each run produces a stable citation rate (cited/total) and run-over-run trend delta (deltaᵣec) that are directly comparable across time, platform, and analyst. A competitive comparison table in the paper shows LLMin8 across nine dimensions vs manual checks, simple trackers (Peec, Mint), and correlation platforms (Profound). LLMin8 is the only approach with: intent-stratified prompt taxonomy, multi-engine coverage, citation quality differentiation (URL vs name mention), a causal attribution pipeline, confidence-graded outputs, Revenue-at-Risk output, and audit trail with reproducibility. The protocol is the data collection layer feeding the LLMin8 LLM Exposure Index (WP-04) and ultimately the Minimum Defensible Causal pipeline (WP-01). Relevant to: GEO tracking, AI brand monitoring, LLM visibility measurement, B2B marketing operations, generative engine optimisation.
LLMin8 Labs (Wed,) studied this question.