Measure the whole path from crawl to qualified lead.

Proposed frozen and live panels

Before the first outcome report, the English protocol will preregister a proposed 60-query frozen panel and a 20-query rotating discovery panel across branded, commercial, informational, comparison, local, trust, pricing/process, and adversarial cohorts.

• Five-observation pilot across different days
• Engine, interface, locale, geography, login state, and page version retained
• Native-authored panels per locale
• Sample sizes updated from observed variance

Separate metrics

• crawl and index eligibility
• AI-answer activation
• retrieval or citation presence
• citation position
• correct entity and service representation
• citation support and completeness
• misattribution and factual error
• referral and engaged session
• qualified lead and conversion

Evidence and reporting rules

• Freeze a query panel and record a separate rotating discovery panel.
• Record engine, product, interface, locale, geography, login state, time, cited URLs, and deployed page version.
• Repeat observations across different days and paraphrases before interpreting a result.
• Separate owner attestation, public-source verification, direct measurement, and planned work.
• Publish negative findings, uncertainty, denominators, limitations, and corrections.
• Use first-party webmaster data, approved APIs, or documented manual sampling.
• Do not treat repeated runs of one query as independent query diversity.
• Suppress public cohorts with fewer than five observations.
• Never publish raw leads or potentially identifying queries.
• Do not claim rank, citation, traffic, or conversion guarantees.
• No prohibited consumer search-result scraping
• No fabricated prompts, citations, reviews, locations, links, or customer outcomes
• No hidden crawler-only facts or keyword variations
• No claim that llms.txt is a ranking factor

Current claim registry

Current evidence references