Skip to content
Start free in Cloud

AI crawler logs and AI referral traffic answer different questions. HitKeep keeps them separate so SEO teams do not mistake crawler activity for demand.

Browser tracking can see many AI-referred human visits. It cannot reliably see AI crawlers because those requests usually do not run JavaScript.

Signal What it means How HitKeep gets it
AI crawler fetch An AI system requested a page or asset Server-side AI fetch ingest
AI-referred visit A human arrived from an AI assistant referrer Browser tracking with hk.js
Correlation A fetched path later received AI-referred visits AI Visibility correlation report

GPTBot fetching /pricing does not mean a buyer visited /pricing. It means an AI system requested the page.

Repeated fetches can show which pages AI systems inspect, where they see errors, and which sections of a site are visible to AI systems. But fetches should not be counted as sessions, users, or conversions.

Question Best signal
Which pages do AI systems request? AI crawler fetch records
Which AI assistants send human visits? AI-referred browser visits
Which fetched pages later receive visits? Correlation report
Where do AI crawlers hit errors? Fetch status code breakdown
Which pages convert after AI referrals? Goals, funnels, and ecommerce reports

Why AI referrals are incomplete by themselves

Section titled “Why AI referrals are incomplete by themselves”

AI-referred visits show human demand, but they miss the discovery layer. If ChatGPT sends five visits to a page, you still may not know whether AI crawlers are also fetching related pages, hitting errors, or ignoring important content.

  • Server-side fetch records show crawler visibility.
  • Browser pageviews show AI-referred human visits.
  • Correlation shows where the two overlap.
  • Discovery: which AI systems fetched the site and which paths they requested.
  • Demand: which pages received AI-referred human visits.
  • Work list: pages with high fetch interest, low referrals, or crawler errors.

This gives the client a practical next step without claiming deterministic attribution, rankings, citations, or guaranteed AI Search visibility.

A useful takeaway is specific and testable: GPTBot and PerplexityBot fetched the comparison cluster, but three high-interest pages returned 404s and none of those paths received AI-referred visits during the period. The next work item is to fix the missing URLs, add internal links from the working comparison pages, and review again after the next crawl window.