Guides/Measuring AI traffic and AI-driven work

Measuring AI traffic and AI-driven work

Two different things get called AI analytics. One is measuring the traffic AI assistants send you, which is a tracking problem because most of it arrives with no referrer and lands in direct. The other is using AI to do the analysis. These guides cover both, and are explicit about which is which.

14 guides in 3 parts

Part 01

Traffic from AI assistants

Where it comes from, why it hides in direct traffic, and how to separate real people from crawlers.

  1. 01
    How to Track ChatGPT Traffic: Measuring AI Assistant Referrals

    AI assistant referrals mostly land in direct or (not set) because desktop apps send no referrer.

  2. 02
    AI Crawlers vs AI Referrals: GPTBot, ClaudeBot and What to Actually Block

    GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, PerplexityBot, Google-Extended, CCBot and Bytespider explained.

  3. 03
    LLM Visibility vs LLM Acquisition: Which AI Search Metric Actually Matters

    LLM visibility tools tell you whether a model mentions your brand. LLM acquisition tells you whether a real person arrived and converted.

  4. 04
    Analytics for AI Agents: Measuring Traffic That Acts on Behalf of a Human

    Session-based analytics breaks completely for AI agents. How to separate humans, agents acting for a user, and crawlers, and what to measure instead: person-level events, declared intent, and task completion.

Part 02

AI doing the analysis

What it automates well, where it is confidently wrong, and how to check it.

  1. 01
    Will AI Replace Data Analysts? The Honest 2026 Answer

    AI will not replace data analysts, but it is absorbing the layer where the question is given and the answer is checkable.

  2. 02
    Using AI to Generate SQL Queries: How to Get Accurate Results Without Hallucinations

    Practical guide to using LLMs for SQL generation covering schema hallucination, join errors, prompting strategies, validation workflows, and when to trust AI versus write queries manually.

  3. 03
    AI in Analytics: Anomaly Detection, Predictions, and Automated Insights

    A comprehensive guide to AI-powered analytics covering anomaly detection, predictive analytics, automated insights, and churn prediction.

  4. 04
    AI-Generated Analytics Reports: Building Workflows That Write Themselves

    How to use AI agents to automatically generate narrative analytics reports from raw data.

Part 03

Agentic workflows

Analytics wired into systems that act on it without a person in the loop.

  1. 01
    AI Agentic Workflows for Analytics: A Practical Guide

    AI agentic workflows let an agent read behavioral analytics and act: detect funnel drops, trigger campaigns, update segments.

  2. 02
    Building an AI Agent Pipeline: From KISSmetrics Data to CRM Actions

    Learn how to build an AI agent pipeline that reads KISSmetrics behavioral data and writes actions into your CRM.

  3. 03
    AI Lead Scoring: Building a Model That Learns

    AI lead scoring learns which behaviors actually predict conversion in your business.

  4. 04
    AI Content Personalization: How to Build the Workflow

    AI content personalization adapts headlines, CTAs, and offers per visitor from behavioral signals.

  5. 05
    GTM Workflow Orchestration: Coordinating Sales, Marketing, and Product Data

    A framework for orchestrating go-to-market workflows across teams and tools.

  6. 06
    The Modern GTM Stack: Workflow Architecture for Analytics-Driven Teams

    A comprehensive guide to designing the workflow architecture of a modern go-to-market stack with analytics at the center.

Common questions

Why does ChatGPT traffic show up as direct in Google Analytics?
The ChatGPT desktop and mobile apps do not send a referrer header, so the visit arrives with no source and GA4 files it under direct or (not set). Only the web version at chatgpt.com reliably passes a referrer.
Is AI crawler traffic the same as AI referral traffic?
No, and they need opposite handling. Crawlers like GPTBot and ClaudeBot fetch pages to train or ground a model and mostly do not execute JavaScript, so they appear in server logs but not in client-side analytics. AI referrals are real humans who clicked through from a conversation.
Should I block AI crawlers in robots.txt?
It is a trade. Blocking training crawlers keeps your content out of model weights, but the same or related agents are often what surface you as a cited source, and citations send real people. Decide per crawler rather than blocking the category.