Your team can already measure AI traffic. The question now is different — out of all those numbers, which ones matter to you? Here are the three families of metrics, the decision each one enables, and what to stop reporting.
Marketing reports now come with sessions, rankings, and "something about AI". The stack to measure it exists, and your team can build it with the guides on this blog — starting with why your AI traffic shows up as Direct. This post doesn't teach you to implement anything: it's the map for reading what that stack produces.
Why Metrics Inherited From SEO Fall Short
AI traffic is different in origin: it arrives from answers, not from lists of links. According to Search Engine Land, pages are now "summarized, excerpted, and cited in environments where clicks are optional". Metrics built to count positions and clicks see none of that.
It's also different in signal — it depends on a fragile referrer that often never arrives — and in behavior. According to Adobe, during the 2025 holiday season, AI-referred visitors were 33% less likely to leave immediately and spent 45% more time on-site than the rest of the traffic. Whoever clicks through from an assistant already talked to the AI first — they arrive better informed.
That's why presence and traffic are two different questions. Rankings measure your place on a list the AI no longer shows; GA4 measures the visits the answer decided to send you. The three families that follow answer three executive questions — and each comes from a different layer of the stack.
Family 1 — Presence: Do You Exist in the Answers?
The executive question is binary — when AI answers for your category, do you show up or not? More and more demand starts inside an answer: Gartner predicted that traditional search engine volume would drop 25% by 2026 due to AI chatbots and other virtual agents.
The metrics to ask for here are three:
- Share of voice — what percentage of your category's answers mention your brand versus competitors.
- Citation rate — how often the AI cites your site as a source in its answers.
- Prompt coverage — how many of the questions that matter to your business you appear in.
All three come from multi-LLM monitoring — systematically querying large language models (LLMs) and measuring who shows up in their answers. These aren't exotic metrics: Search Engine Land lists them among the eight GEO (generative engine optimization) metrics it recommends tracking in 2026.
The complement is the AI Visibility Score, which doesn't measure whether you appear but how ready your site is to be cited — 130+ technical check items. Presence is the outcome; readiness is the cause you actually control.
Ask for these metrics as a monthly series, not a one-off snapshot. Share of voice moves slowly — the executive signal is in the change between snapshots, not in any single day's value.
The decision it enables: where to invest in content and authority — and what to fix on your site before expecting results.
Family 2 — Acquisition: How Much Arrives, and From Where?
Here the question is about the channel: how much AI traffic arrives, which sources drive it, and is it growing? The metrics: AI traffic share over total sessions, source mix — ChatGPT, Gemini, Claude, Perplexity — and month-over-month trend.
They come from Google Analytics 4 (GA4) in two layers that coexist. The native "AI Assistant" channel is the zero-maintenance floor; the custom AI traffic channel adds control over the source list. Your team can build both with those guides — no need to choose.
The caveat worth internalizing: the number you see is a floor, not the total. Part of AI traffic arrives without a referrer due to referrer stripping and hides in "Direct". Server-side tracking recovers part of that signal — but when reading the report, treat your AI share as a confirmed minimum.
The context to calibrate expectations: according to Similarweb, AI-referred traffic remains below 1% of global internet traffic. Small in volume and growing fast — Adobe measured a 693% year-over-year increase in AI traffic to US retail sites during the holiday season. That contrast — small volume, high growth — is exactly what a monthly trend captures and a yearly snapshot hides.
The decision it enables: which sources to prioritize and whether the channel deserves more attention. With today's share still small, the actionable number isn't the absolute — it's your trend, month against month, against yourself.
Family 3 — Value: What Is That Traffic Worth?
The third question is the budget question: does that channel deserve investment? The metrics: conversion rate (CR) of AI traffic against your other channels, value per session, and assisted conversions.
The mechanism to compare already exists in your stack: AI session audiences in GA4 put the conversion rate of AI traffic next to that of organic or paid, in the same reports your team already uses. Assisted conversions complete the picture: AI also pushes decisions mid-journey, not just at the last click.
The public evidence points in a consistent direction: AI traffic converts differently because it arrives further down the funnel. According to Semrush research reported by MarTech, the average AI search visitor is 4.4 times as valuable as the average visit from traditional organic search, measured by conversion rate. And Adobe reported that during that same holiday season, AI-referred traffic converted 31% more than other traffic sources.
These are cited third-party benchmarks, not promises — real value varies by industry and by business, and yours is measured with your own data. How to translate that value into currency for the board report deserves its own analysis — coming soon.
The decision it enables: how much budget the channel deserves — defended with conversion and value per session, not with volume.
The Three Families in a Single View
Table 1 — the metrics of the three families in one executive view. The last column, "The decision it enables", is the one you take to the committee.
| Metric | What it answers | Where it comes from | The decision it enables |
|---|---|---|---|
| Share of voice in AI answers | How often do I appear versus competitors? | Multi-LLM monitoring | Where to invest in content and authority |
| Citation rate | How often am I cited as a source? | Multi-LLM monitoring | Which content deserves reinforcement |
| AI Visibility Score | How ready is my site to be cited? | Technical site analysis | What to fix before investing in presence |
| AI traffic share + source mix | How much arrives, and from where? | GA4 — native + custom channel | Which sources to prioritize |
| AI channel trend, month over month | Is the channel growing or stalling? | GA4 — same channel, monthly view | Whether the channel deserves more investment |
| AI traffic CR vs. other channels | Does it convert better or worse than the rest? | GA4 audiences | How much budget it deserves |
| Value per session | How much does each AI visit contribute? | Audiences + revenue data | The business case for the board |
If your monthly report carries these seven rows, it carries what you need to decide on the AI channel. Everything else — configuration, regex, events — is the stack's job, not the committee's.
Anti-Metrics: What to Stop Accepting in Reports
Just as important as asking for the right metrics is refusing the ones that enable no decision. Four patterns we see in the wild:
- Total raw sessions with no source breakdown. The AI channel disappears inside the aggregate — it could have doubled or collapsed and the total won't tell you.
- SEO rankings as a proxy for AI visibility. Ranking on Google doesn't mean being cited by ChatGPT — presence is measured in the answers, with family 1's metrics.
- "Direct traffic is up" celebrated without questions. Part of that Direct may be AI traffic in disguise due to referrer stripping — before celebrating, ask how much.
- AI "appearances" with no citable source or method. If no one can audit how it was measured — which prompts, which models, which dates — it doesn't go to the board.
The point isn't that your current report is wrong. It's that AI traffic needs its own metrics — inheriting SEO's leaves the three executive questions unanswered.
Frequently Asked Questions
What share of my traffic should come from AI?
There's no single benchmark. AI-referred traffic is still a small fraction of the total in almost every industry, though it's growing fast. The actionable move isn't comparing yourself to a magic number — it's tracking your own month-over-month trend.
Why does my AI traffic look so small in GA4?
Because part of AI traffic arrives without a referrer and gets classified as Direct. The number you see in any channel — native or custom — is a floor, not the total. Read it as a confirmed minimum and decide with the trend.
If I can only ask for one metric, which one comes first?
It depends on your stage. If you don't yet know whether AI mentions you, start with share of voice and the AI Visibility Score — without presence there's no traffic to measure. If you already show up in answers, move on to AI traffic share and its trend.
Are Google rankings useless now?
They're not — for SEO they still work. What they don't do is serve as a proxy for AI visibility: ranking on Google doesn't mean being cited by ChatGPT. They're two different questions, and each needs its own metrics.
Wrapping Up
Three ideas to take with you:
- Three families, three questions: presence — do I exist in the answers?; acquisition — how much arrives and from where?; value — what is that traffic worth?
- AI traffic numbers in GA4 are floors, not totals — read them as a confirmed minimum and decide with the trend.
- Anti-metrics are a decision too: whatever can't be audited or changes no decision goes out of the report.
Family 1 orders the rest — without presence in the answers there's no traffic to measure and no value to discuss. That starting point is measured with the AI Visibility Score, the 130+ check item engine we run on every site.
Enter a URL and get your AI Visibility Score in under 60 seconds.