Who's actually reading us

The bot charts below cover declared crawlers (Googlebot says it's Googlebot). IP org lookup is the other half: what do the undeclared IPs actually resolve to?

From the last 7 days of Apache logs, roughly 38% of unique IPs resolve to cloud hosting ranges (OVH, AWS, DigitalOcean) — these are content scrapers pulling pages for training data or competitive indexing, not humans. About 8% hit residential proxy pools operated by rank-tracking services (Datacamp, Code200, M247): they send Google a search query to generate the GSC impression, then fetch the actual page through a proxy to confirm the ranking — which is why GSC impressions can spike on a page with barely any real clicks.

The 21% residential ISP traffic (Bell Canada / Sympatico leads the list) is the one category that could contain real people, though some of it is also the operator's own traffic before a DHCP lease change got filtered in GA4. The remaining 34% is unclassified — smaller regional ISPs, hosting blocks ip-api.com doesn't tag clearly. Weekly Haiku analysis runs Sunday mornings and writes the breakdown to a private cache file used by this section.