01 · Define
Start with decisions, not vanity prompts.
Define one product, market, or growth journey. Build a prompt library that represents how a prospective customer might discover, compare, qualify, and choose an institution.
Separate branded prompts from unbranded prompts. Record location, product constraints, audience, and intent so later runs compare like with like.
02 · Observe
Capture answers as time-bound observations.
Run the same prompt library across the selected answer environments. Record the date, prompt, response, brands named, position and framing, citations, and material factual errors.
An answer is not a permanent ranking. Results can change by model, mode, location, personalization, and time. The benchmark is a repeatable snapshot, not a universal truth.
Core measurement dimensions
Answer visibility
Whether the institution appears for relevant unbranded questions, and in what context.
Citation share
Which sources are cited and how often the institution’s owned or corroborating sources appear.
Fact accuracy
Whether products, eligibility, locations, and other decision facts are represented correctly.
Competitive position
Which alternatives are named, why they appear, and which evidence supports them.
Journey readiness
Whether a person can move from the answer to a clear, trustworthy, and compliant next step.
03 · Diagnose
Trace each gap to a fixable signal.
Review the institution's product pages, branch and service-area facts, structured data, internal linking, third-party corroboration, reputation signals, and conversion path.
Prioritize work by customer importance, evidence strength, implementation effort, compliance requirements, and the number of prompts or journeys it can improve.
04 · Re-measure
Prove movement without overstating causality.
After approved changes are published and crawlable, re-run the controlled prompt set. Compare visibility, citations, fact accuracy, competitive context, and journey quality against the baseline.
Report what changed, what did not, and what remains uncertain. Pair answer observations with aggregate search, referral, engagement, and conversion signals where available.