GA4 Now Has an AI Channel. Here's How to Report AI Visibility to Leadership Without Fooling Yourself

Short answer: since May 13, 2026, Google Analytics 4 groups visits from recognized AI assistants into a new "AI Assistant" channel. It is a useful start, but it undercounts: the same AI source can be split across several channels, many visits arrive without referrer data, and most AI influence never produces a click at all. A credible leadership report combines three layers: sampled visibility in AI answers by engine, first-party citation data where it exists, and traffic and conversions from a custom GA4 channel, presented as trends with clear caveats.

What GA4 added

On May 13, 2026, Google updated GA4's default channel grouping. When GA4 detects a referrer from a recognized AI assistant such as ChatGPT or Gemini, it sets the medium to "ai-assistant," the campaign to "(ai-assistant)" and the channel to "AI Assistant," so AI traffic appears next to organic search in standard reports (PPC Land). "When Google Analytics detects a referrer that matches a recognized AI assistant, it automatically assigns a new value - 'ai-assistant' - to the medium dimension." - PPC Land (Google Analytics Help Center, 2026) - product update.

For the first time, many teams can show AI-referred traffic without building anything. That makes it tempting to report the new channel as "our AI performance." It isn't.

What the AI Assistant channel misses

An analysis by Itamar Blauer for Search Engine Journal found three problems with the default channel (Search Engine Journal). First, one source can be split: sessions from chatgpt.com showed up in the AI Assistant, Referral and Unassigned channels at the same time, because GA4 matches on source and medium together. Second, the list of recognized assistants changes; Claude was removed from it between launch and June 2026, and Perplexity traffic was landing in Referral. Third, ChatGPT's mobile app and embedded browsers often strip referrer data, so some visits arrive with no AI attribution at all. "One source, scattered across three channels." - Search Engine Journal (Itamar Blauer, 2026) - analysis.

The bigger gap is not technical. Many buyers read an AI answer, form a shortlist and later visit a vendor directly or search its name. That influence shows up, if at all, as direct or branded search traffic. Any report built only on AI referrals measures the smallest visible part of AI's effect.

A three-layer reporting framework

Layer 1: Visibility in AI answers

This is the leading indicator. Track how often your brand is mentioned and cited for a fixed set of buyer questions, by engine and by query cluster, sampled across many runs. Report it as a rate over time, not a rank. Keep the question set stable and documented so changes reflect the market rather than your sample. Some GEO providers, including Hordus AI, report this kind of answer share against a fixed tracked question set with the methodology alongside the figures (Hordus); whoever runs it, ask for the same transparency.

One new caveat belongs here. A September 2026 study of 8,609 AI responses found that a user's history can change which brands are recommended, with the largest effect on Gemini and none detected on Claude, and that history-free tracking recovered about 79% of the brands seen in personalized sessions (EIN Presswire). It is a single small study, but it is a reason to present tracked visibility as an estimate of what buyers see, not an exact measure.

Layer 2: First-party citation data

Where an engine provides its own data, use it. Bing Webmaster Tools' AI Performance report, launched in February 2026, shows how often a site is cited in Microsoft Copilot and Bing's AI summaries and which pages are referenced (Bing). It covers one ecosystem, but it is measured data rather than a sample.

Layer 3: Traffic and outcomes

This is the lagging indicator. Use GA4 for sessions, engagement and conversions from AI sources, but through a custom channel rather than the default one, and add the signals that capture influence without clicks: answers to "How did you hear about us?" that include AI assistants, and trends in branded search and direct traffic for the products you are working on.

Building a custom AI channel in GA4

The fix Blauer recommends is a custom channel group that matches on source alone, ignoring medium:

• Create a custom channel group in GA4's admin settings, starting from a copy of the default group.
• Add an "AI" channel that matches source against a list of AI domains, such as chatgpt.com, perplexity.ai, gemini.google.com, claude.ai and copilot.microsoft.com, and maintain that list yourself.
• Place it above Referral so AI sources are captured before they fall into the general referral bucket.
• Use it across your full date range: Custom channel groups apply to historical data, which gives you a consistent trend line.
• Check Unassigned and Direct regularly for patterns that suggest missing AI traffic, especially from mobile.

What to show leadership

Executives need a one-page view, not a dashboard tour. A useful monthly report has five parts: visibility rate by engine for your priority question clusters, with the trend; citations from first-party sources such as Bing; AI-referred sessions and conversions from the custom channel; self-reported AI influence from forms and sales notes; and two or three real examples of what AI assistants currently say about you and your competitors. Add one line on what the data cannot show.

Example: a B2B software company reports that its mention rate in Perplexity and AI Overviews rose from 18% to 31% over a quarter for its main category questions, while ChatGPT stayed flat. AI-referred sessions grew modestly, but demo requests citing an AI assistant doubled. Leadership sees a coherent story: visibility is rising on search-based engines, it is starting to show up in pipeline, and ChatGPT is the next gap to close.

Common reporting mistakes

• Reporting the default GA4 channel as the whole picture: It undercounts by design.
• Blending all engines into one score: Engines cite different sources, so averages hide where you're absent.
• Showing an "AI rank": AI answers vary run to run, so positions are not meaningful.
• Judging month to month: Visibility is noisy; quarterly trends are more reliable.
• Hiding the caveats: Stating what the data can't capture builds more trust than overclaiming.

The bottom line

GA4's AI Assistant channel makes AI traffic visible, but traffic is the smallest part of AI's influence on buyers. Reporting that leadership can trust combines what AI engines say, what first-party data confirms and what shows up in traffic and pipeline, presented as trends with honest limits.