Brantial
Get Audit

PLATFORM / AGENT ANALYTICS

AI Agents Are on Your Site — See Who Arrives, at Log Level

GA4 either misses agent traffic or files it under 'other'. Agent Analytics classifies 27 bot signatures at log level — GPTBot, ClaudeBot, PerplexityBot and Google-Extended included — shows which of your pages feeds which model, and ties an agent visit to the human click that follows.

PRODUCT VIEWFrom the live product interface
Bot VisitsBy agent, 7 days
GPTBotClaudeBotPerplexityBot
All Bot VisitsLast 30 days
32.6K+12K
27 SIGNATURES · LOG LEVEL
Agent SplitRequest share, 7 days
GPTBot4.8K
PerplexityBot3.2K
ClaudeBot2.0K
Google-Ext.1.4K
The Attribution ChainFrom crawl to click
Agent visit loggedOK
Citation found in answerOK
Human click matchedRUN
Page value reported

What Is Agent Analytics?

GA4 either misses agent traffic or files it under 'other'. Agent Analytics classifies 27 bot signatures at log level — GPTBot, ClaudeBot, PerplexityBot and Google-Extended included — shows which of your pages feeds which model, and ties an agent visit to the human click that follows.

Figures come from the sample account; your console shows them in real time.

Seven Days of Agent Traffic, Sample Account
  • GPTBot: 4.812 — /blog/geo-rehberi — +18%
  • PerplexityBot: 3.204 — /fiyatlandirma — +9%
  • ClaudeBot: 1.988 — /docs — +31%
  • Google-Extended: 1.412 — /urunler — -4%
  • Unknown (high heat): 356 — /api-docs — New
Crawl Budget Is Now Answer Budget
  • The Leading Indicator: If a model cannot crawl your site, it cannot cite it; agent traffic is the leading indicator of visibility.
  • Content Value Signal: Seeing which pages get crawled shows which content models consider worth reading.
  • A True Bot Inventory: Impersonator filtering also hands your security team a true bot inventory.
  • Page-level ROI: The attribution chain answers 'does AI traffic pay' at page level.
The Modules That Feed and Use Agent Data
  • Site Auditor: Finds the root cause of uncrawled pages: robots, schema, access.
  • Integrations: Cloudflare, Vercel and GA4 connect in five minutes.
  • Prompt Volumes: Matches crawled pages to the queries with demand.

From Logs to an Attribution Chain in Four Steps

Setup takes one of two routes: edge log forwarding or a one-line JS snippet.

Collect

Your Cloudflare, Vercel or nginx logs stream in; without log access, a lightweight JS snippet captures server-side hits.

Classify

Requests are split across 27 known agents by UA, IP range and behaviour signature; impersonators are filtered and unknown agents get a heat score.

Attribute

The chain agent visit → citation inside an answer → human click is built, showing which crawl produced real traffic.

Report

A per-page matrix of which model each page feeds, so robots and llms.txt decisions are made with data.