{
 "window_days": 14,
 "generated": "2026-08-20T10:02:05.183Z",
 "totals": {
  "human": 5093,
  "agent": 415,
  "search": 44,
  "bot": 37,
  "unknown": 3,
  "tool": 72
 },
 "reads": 5592,
 "agent_share_percent": 7.4,
 "by_agent": {
  "claudebot": 308,
  "oai-searchbot": 10,
  "ccbot": 6,
  "chatgpt-user": 65,
  "gptbot": 9,
  "perplexitybot": 17
 },
 "by_referrer": {
  "www.reddit.com": 2
 },
 "days": [
  {
   "date": "2026-08-20",
   "agent": 13,
   "bot": 3,
   "human": 230,
   "tool": 50
  },
  {
   "date": "2026-08-19",
   "agent": 108,
   "bot": 34,
   "human": 2992,
   "search": 44,
   "tool": 22,
   "unknown": 2
  },
  {
   "date": "2026-08-18",
   "agent": 45,
   "human": 1507,
   "unknown": 1
  },
  {
   "date": "2026-08-17",
   "agent": 249,
   "human": 364
  }
 ],
 "capped": false,
 "note": "Content reads only — assets, images and API calls are excluded, so this counts readers rather than HTTP requests. No IP, no cookies, no per-visitor data is stored. Each read is written as its own key, so no event is lost to a lost increment. \"capped\" means the key listing hit its limit and the figure is a lower bound. High-volume classes (human, tool) are sampled at 1:10 and multiplied back, so those two are estimates rather than exact counts; agent, search and bot stay one key per event. \"tool\" counts non-reader automation (node, headless browsers, shells) and is excluded from \"reads\" and from the agent share; own monitoring is not counted at all."
}