For thirty days the Observatory turned its instruments on its own property. The intent was simple: produce a sample bulletin against a subject we know in detail, and let prospective applicants judge what the form looks like in practice. The subject is small. The activity directed at it was not.

The site received attention from a recognizable set of declared crawlers — among them Apple's Applebot, declared AI crawlers from established model labs, and the usual mainstream search indexers. None of this is unusual. The texture was. Declared crawlers behaved at patterns inconsistent with their public declarations, and several appeared to coordinate around the same surfaces in narrow windows.

More notably, the property attracted persistent, non-interactive activity from infrastructure that does not declare itself — a rotating residential proxy network with global reach, an industrial automation provider transiting Chinese hosting, and the characteristic patterns of agentic browser frameworks reading the site on behalf of unknown principals. None of this is "attack" traffic. All of it is observation.

The cohort context matters. Three of the actors observed against botconduct.org appear in current bulletins for other cohort properties — including properties in adjacent verticals. The shared interest is not random.

The reason the cohort matters is simple: most actors directed at a property are not only directed at that property. The same operator, the same framework, the same infrastructure shows up against many sites in parallel — sometimes with intent that becomes legible only when the pattern is observed across the cohort.

The Observatory's contribution is not the detection of any single visit. The contribution is the placement of that visit inside a wider pattern that the receiver, looking alone, cannot see. A single property looking at its own logs sees noise. The Observatory sees a route.

For the present subject, three of the four characterized actors observed against botconduct.org are currently active against other cohort properties. The framework signature is shared. The residential proxy signature is shared. The industrial hosting signature is shared. The operational pattern appears across properties consistently.

This is the structural argument for receiver-side observation, viewed across a curated cohort, rather than per-property bot management read in isolation. The signal each property sees is partial; the signal the cohort sees is interpretable.

The BotConduct Observatory operates as a closed oracle. We disclose findings, not method. This is a deliberate choice: it protects both the integrity of receiver-side observation against operator adaptation, and the long-term value of the cohort to its members. Approved members receive interpretation. They do not receive the playbook.