Browse documentation

Agent Analytics

AI bots like GPTBot visit your website to read your content. Agent Analytics shows you which bots came, what they read, and whether that reading actually turned into your brand being recommended in AI answers.

Know which AI systems are reading your site, what pages they access, and whether that crawl activity converts into actual visibility in AI answers.

The Two Sides of GEO

Most GEO tools measure only the output side: what AI engines say about you. A2Z Reach also measures the input side: which AI crawlers visit your site and what they read.

Connecting both sides reveals the most powerful insight in GEO: the conversion gap. If GPTBot crawls your pricing page 50 times per week but ChatGPT never recommends your pricing as competitive — something is wrong with how your content is structured for AI consumption.

What We Track

  • AI crawlers identified: GPTBot (OpenAI training), OAI-SearchBot (OpenAI indexing), ChatGPT-User (live user fetch), PerplexityBot, ClaudeBot, Google-Extended, Googlebot (AI Overviews), and more
  • Crawl volume: How many requests per bot per day, with trends
  • Pages accessed: Which content pages bots read (filtered — no static assets noise)
  • Purpose classification: Training vs indexing vs live-user citation fetch
  • Cache behavior: Are bots getting fresh content or stale cached versions?
  • Blocked/errored pages: Content you have that bots can't read (a GEO problem)
  • AI-referred traffic: Humans clicking through from AI answers to your site

How to Connect

Agent Analytics ingests your CDN access logs via a webhook. Supported CDN providers:

  • CloudFront (real-time logs via Firehose)
  • Cloudflare (Logpush or Worker)
  • Fastly (real-time log streaming)
  • Akamai (DataStream2)
  • Google Cloud CDN (Log Router)

Setup takes 10–15 minutes. No code changes to your application — you're routing existing CDN logs to our ingest endpoint.

Bot Verification

Not every request claiming to be GPTBot is real. Bots can be spoofed. A2Z Reach verifies crawler authenticity using:

  • IP range validation: Cross-referencing source IPs against published bot ranges (OpenAI, Google, etc.)
  • Reverse DNS: Confirming hostnames match the claimed bot identity

Unverified requests are flagged separately — they still count for volume but don't pollute your verified crawl analytics.

AI Referral Tracking

When a human clicks a link from an AI answer (e.g., ChatGPT cites your page and the user clicks through), that appears in your CDN logs with a Referer header from the AI platform. We classify these as AI-referred traffic — the closest thing to a "conversion" in GEO.

Attribution tiers:

  • Tier 1 — Referer: The HTTP Referer header shows chatgpt.com, perplexity.ai, etc.
  • Tier 2 — UTM: utm_source=chatgpt.com (how ChatGPT mobile links work)
  • Google (blended): Google Search and AI Overviews share the same referer — labeled as "Google (Search + AI Overview)" since they can't be distinguished in logs

The Content Pages View

The pages tab shows which of your content pages AI bots are reading — filtered to real pages only. Static assets (JS, CSS, images, fonts), API paths, and security probes are automatically excluded. GEO-relevant discovery files (robots.txt, sitemap.xml) are always shown.

Pages are classified as:

  • Served: Bots got a 200 response — your content is accessible ✓
  • Blocked: Real content that some bots couldn't read (returned 4xx/5xx) — a GEO problem to fix