Live, public traffic and retrieval signal for bifrostindex. Refreshed about every 15 minutes. We publish this so practitioners and AI assistants can see what's actually being read, not just what we claim to cover.
Last 30 days
37,913 public requests · 79 live user fetches · 14 AI reads of the top guide · 96 AI search crawler fetches · 1,646 training-crawler visits · 3,276 browser-shaped visits.
All time · since May 24, 2026
Model-building crawls (GPTBot, ClaudeBot, Google-Extended, and others). High volume, low intent — a training crawler fetches every URL on the sitemap whether or not anyone ever asked a question about the topic.
Pre-indexing bots for AI-powered search engines (OAI-SearchBot, PerplexityBot, Claude-SearchBot). Analogous to Googlebot — they index URLs so future answers can cite them. Higher intent than training crawlers but not tied to a live user question.
An AI assistant fetching content while answering a user's question right now (ChatGPT-User, Perplexity-User, Claude-User). The highest-intent signal we see — the assistant decided this specific URL answers the person's question.
Traffic that could be a human reader — not a known bot, not an operator session, not a high-fanout scraper. May include AI-assisted browsing or automated agents with browser-like UAs, so the real-human share is smaller than this number.
Traditional search-engine bots, Common Crawl, and other uncategorized automated clients — distinct from AI retrieval and AI training traffic.
The scale of each signal, from raw request volume down to the live user fetches that put a guide in front of an AI assistant's user right now. Bars are sized against total public requests. The four AI signals are parallel, not nested — but guide reads are a subset of live user fetches.
Every public pageview in the window.
GPTBot, ClaudeBot, Google-Extended, and others — model-building fetches.
OAI-SearchBot, PerplexityBot, Claude-SearchBot — pre-indexing for AI-powered search engines.
ChatGPT-User, Perplexity-User, Claude-User fetching content while answering a question right now.
Public pages only — excludes admin, internal dashboards, API calls, sitemaps, and robots.txt. 37,913 pageviews in the last 30 days. AI retrieval is shown as two separate rows: live user fetches (a person asked a question right now) and AI search crawlers (pre-indexing). The real-human share is inside the browser-shaped bucket but smaller than the number shown.
Same 37,913 pageviews as above, grouped by page type. Guide pages is the only bucket that maps to our editorial product — the rest is home page, browse/index pages, marketing & legal, and legacy /questions URLs (308-redirect now). AI columns: live = user-driven fetches; idx = AI search crawler pre-indexing.
All public paths fetched by any ai_search bot in the last 30 days — both live user fetches (ChatGPT-User, Perplexity-User, Claude-User) and AI search crawlers (OAI-SearchBot, PerplexityBot, Claude-SearchBot). The per-row bot breakdown shows which bots hit each path. 166 content retrievals across 42 paths. Scanner probes (9 hits) are separated below.
The three distinct kinds of AI traffic: live user fetches (highest intent), AI search crawlers (pre-indexing), and training crawlers (model building). Bars are scaled independently within each group.
ChatGPT-User, Perplexity-User, Claude-User — a real person asked a question right now.
OAI-SearchBot, PerplexityBot, Claude-SearchBot — pre-indexing for AI-powered search (like Googlebot for AI).
Traditional search-engine bots (Googlebot, Bingbot, …), Common Crawl, and other uncategorized automated clients. Excludes humans and the AI traffic broken out above.
Top 25 guide URLs by total requests, out of 15,231 guide-page hits in the last 30 days. The bar under each row shows the source mix for that page. Pink = live user fetches · orange = AI search crawlers · purple = training crawls · cyan = browser-shaped.
Once human contributors have a public confirmation feed, this section will rank them by contribution frequency and contribution type — sections drafted, sections confirmed, edits suggested, sources cited. For now, the editorial author is the BifröstIndex bot, with all sections marked “not yet reviewed by a human” per the publication disclosure on each guide.
Subset of live user fetches that landed on an editorial guide page — the closest signal to a live user reading our content via AI.
A User-Agent can be spoofed, so we list destinations exactly as observed rather than filtering them. The signal that matters is the guide and browse URLs near the top.
AI-labelled clients also scan for files that don't exist on a content site (e.g. /.env, /.git/config, /actuator/env). Excluded from the headline count above but disclosed here.
GPTBot, ClaudeBot, Google-Extended & others — model-building crawls. High volume, lowest intent.
