Zero-click: what GA4 shows about AI traffic and what it misses

Since May 13, 2026, Google Analytics has a dedicated "AI Assistant" channel in the Default Channel Group. According to Google's "What's new" page, a visit is assigned to it when the referrer matches a recognized AI assistant. The announcement names ChatGPT, Gemini and Claude as examples. Google's channel definition lists ChatGPT, Gemini, Deepseek, Copilot and Grok as sources, and states that the channel excludes Google's own AI Overviews and AI Mode.

The update changes three dimensions:

Before this, those visits landed in Referral, Direct or Unassigned, depending on the source and the browser. Teams that wanted to see them built a custom channel group with a regex over source names. That workaround still works for sources Google does not recognize, but the default report now covers the main assistants.

ChatGPT adds a second signal. Links it shows to sources carry the parameter utm_source=chatgpt.com, so the visits can be tracked in tools such as Google Analytics. Search Engine Roundtable reported in June 2025 that ChatGPT extended the parameter from cited sources to the links in its "More" section. OpenAI's own help page that describes this parameter was not reachable when we checked on October 13, 2026, so treat the exact wording as secondary and test it in your own reports.

GA4 records a session when someone arrives on your site and your tag fires. That definition leaves out four things:

1. Answers without a visit. A user asks ChatGPT for the best tool in your category, reads three names, and goes to a competitor or to your homepage by typing the URL. GA4 sees at most a Direct visit later, with no link to the answer.
2. Visits without a referrer. Apps, privacy settings and copy-pasted links can strip the referrer. The visit appears as Direct, so the AI channel is a floor, not a total.
3. Google's own AI surfaces. Google's documentation says AI Overviews and AI Mode traffic is counted in the Search Console "Web" search type, and the clicks come to your site as Google organic. In GA4 they sit inside Organic Search, not in the new AI Assistant channel. You cannot read them as a separate line in GA4.
4. Crawls and fetches. AI crawlers do not run your JavaScript tag, so GA4 never sees them, and neither does it see the pages they read to build an answer.

The practical rule: the AI Assistant channel tells you how many people clicked through from an assistant. It does not tell you how many people saw your brand.

Two studies are worth knowing, and both have limits.

• Pew Research Center tracked the browsing of 900 US adults (68,879 Google searches, 12,593 of them with an AI summary). On visits to a results page with an AI summary, people clicked a traditional result 8% of the time, against 15% on pages without one. Only 1% of visits with a summary included a click on a link inside it. The data is from March 2025 and covers Google only. It is observational, so it shows an association, not the size of the effect on your site.
• Ahrefs compared 300,000 keywords (150,000 with an AI Overview, 150,000 informational ones without) using aggregated Search Console data on desktop click-through rate. Its December 2025 update says an AI Overview correlates with a position-one CTR about 58% lower than expected, up from 34.5% in its earlier study. Ahrefs sells SEO tools, and the result is correlational: queries that trigger an overview differ from queries that do not.

Neither study says how many of those users saw your brand name in the answer. That is the number GA4 cannot give you and that no click-based tool can.

GA4 measures people through a JavaScript tag. AI crawlers fetch pages server-side, so they never execute it. OpenAI documents several bots with different jobs:

The settings are independent. A site can allow OAI-SearchBot to appear in search and disallow GPTBot. OpenAI also notes that after a robots.txt change it can take about 24 hours for its search systems to adjust.

For measurement, this means crawler activity lives in server logs or CDN logs, not in GA4. Some vendors sell this as "agent analytics". It answers a different question from GA4: whether engines fetch your pages, not whether people click from them.

A short, practical setup:

1. Use the default AI Assistant channel for the main assistants, and keep a custom channel group only for sources it does not cover.
2. Segment by landing page. AI visits tend to land on specific articles. A growing AI channel on a few pages tells you which content engines cite.
3. Watch Direct alongside AI. If AI Assistant stays flat while Direct and branded search grow after you publish more citable content, assistants may be influencing people without sending a click. Treat that as a hypothesis to test, not a result.
4. Combine GA4 with Search Console. Search Console covers Google's AI features inside Web, GA4 covers assistant referrals, and neither covers a mention that produced no click.
5. Check your crawl access. If OAI-SearchBot is blocked, your site is less likely to be surfaced in ChatGPT search results, so expect fewer referral visits to measure.

Click data answers "how many arrived". Visibility data answers "how often were we named". They are separate measurements, and the second one needs a different tool.

At Humanswith.AI we treat them as three layers:

• Visibility: how often the brand appears in answers to a fixed set of prompts, across engines.
• Citations: which of your pages the engines name as sources, and which sources they cite instead of you.
• Traffic: what GA4 and Search Console already show.

We make a product in this category, so weigh that. Our own audit of 158 publications found articles older than two months cited by AI engines in 43% of cases, against 7% to 15% for articles younger than a month. That is why we re-measure after two months, not two weeks, and why a flat GA4 line in the first weeks says little.

If you are choosing a tool for the visibility layer, our comparison of the market is here: AI visibility tools in 2026: 10 platforms compared with pricing.

Report these side by side. A single "AI traffic" number invites the wrong conclusion in both directions: a small number looks like failure, a growing one looks like proof.

• Google Analytics Help, "What's new in Google Analytics", entry of May 13, 2026 (support.google.com/analytics/answer/9164320)
• Google Analytics Help, "Default channel group" (support.google.com/analytics/answer/9756891)
• Google Search Central, AI features and your website
• Search Engine Roundtable, OpenAI's ChatGPT Updates To Improve Tracking In Analytics (June 2025, third party)
• OpenAI, Overview of OpenAI crawlers
• Pew Research Center, Google users are less likely to click on links when an AI summary appears in the results (July 2025)
• Ahrefs, Update: AI Overviews Reduce Clicks by 58% (third party, correlational)
• Humanswith.AI internal audit of 158 publications, the same figures as in AI visibility tools in 2026