A longitudinal measurement system for studying when brands are mentioned, cited, recommended or omitted across AI discovery systems, and what changes appear to influence those outcomes.
Public events tell us what changed in the ecosystem. The Observatory records what changed in actual outputs, citations and representation over time.
Define repeatable commercial and informational queries around a brand, category or entity.
Capture the answer, platform, model context, citations, retrieved domains and representation signals.
Change one meaningful variable such as entity markup, page structure, source coverage or crawler policy.
Track whether visibility, citation persistence, recommendation status or ambiguity changes.
Experiments are designed to distinguish correlation from stronger causal evidence. Results remain classified as established, strong, emerging, mixed/open or disconfirmed where appropriate.
Test whether publishing and revising llms.txt produces any measurable change in retrieval, citation or representation outcomes.
Test whether clearer machine-readable relationships between an organization, founder, products and canonical pages correlate with more consistent AI representation.
Measure how often the same domains and pages remain cited across repeated runs of the same prompt set.
Customer audits, contributed prompts, URLs and response text remain private by default. Public research uses only explicitly publishable records or anonymized and aggregated findings.
Studies may expose methods, aggregate findings and approved examples where publication rights are clear.
Commercial measurement records are not exposed by the public Observatory unless the contributor explicitly approves publication.
A public observation stream will only launch after record-level publication controls and consent rules exist in the data model.
Repeated observations matter more than isolated screenshots. Probabilistic systems require longitudinal samples.
Prompt wording, platform, model context, geography, language and timing are treated as experimental variables rather than noise.
Public research can remain open while company-specific diagnosis, monitoring and intervention strategy becomes the premium layer.