How To: Track GA4 AI Assistant Traffic, Including Historical Data

TL;DR:
GA4 now has a default AI Assistant channel for recognised traffic from tools such as ChatGPT, Gemini, Claude and Perplexity. You can view this traffic directly in GA4, but the channel doesn’t give you a complete historical picture. To report AI assistant traffic further back, I use Looker Studio calculated fields to combine GA4’s newer AI Assistant classification with older referral traffic from known AI platforms.

What Is the AI Assistant Channel in GA4?

The AI Assistant channel is a default channel group in Google Analytics 4 used to identify traffic from recognised AI assistants and chatbot platforms.

GA4 introduced the channel in May 2026, making it easier to separate identifiable AI assistant traffic from Referral and Direct traffic.

When GA4 recognises a visit as coming from an AI assistant, it can classify the session using:

• Medium: ai-assistant
  
• Default channel group: AI Assistant
  
• Campaign: (ai-assistant)

This can include traffic from platforms such as ChatGPT, Gemini, Perplexity, Claude, Copilot, DeepSeek and Grok.

However, that doesn’t mean GA4 will identify every visit originating from an AI platform. The referrer needs to be available and recognised by Google for the session to be classified as AI Assistant.

That’s particularly important when looking at historical data. Before the AI Assistant channel was introduced, identifiable traffic from these platforms was generally classified elsewhere in GA4, commonly as Referral traffic.

So there are really two parts to reporting AI traffic in GA4:

1. Reporting traffic GA4 now classifies as AI Assistant
   
2. Identifying historical AI referral traffic from before the channel existed

I’ll cover both below.

How To Create An AI Assistant Traffic Report In GA4

The simplest way to see AI Assistant traffic in GA4 is through the standard Traffic acquisition report.

Go to:

Reports > Acquisition > Traffic acquisition

Then set the primary dimension to:

Session default channel group

From there, search or filter for:

AI Assistant

This will show sessions that GA4 has recognised and classified as AI Assistant traffic.

You can report on metrics such as:

• sessions
  
• users
  
• engagement rate
  
• key events
  
• revenue

For most reporting, I’d start with Traffic acquisition because it shows session-level acquisition. You can also use User acquisition if you specifically want to understand which channels originally acquired users.

This is how AI Assistant traffic currently appears in my own GA4 account:

The native report makes ongoing AI traffic much easier to monitor. However, there’s an obvious limitation if you want to analyse a longer period: the AI Assistant channel was only introduced in May 2026.

That brings us to the historical reporting problem.

Can GA4 Report Historical AI Assistant Traffic?

Not directly through the new AI Assistant channel.

GA4 only introduced AI Assistant as a default channel in May 2026. Before that, identifiable traffic from platforms such as ChatGPT, Claude, Gemini, Perplexity and Copilot could appear under other channels, most commonly Referral.

If no usable referrer was passed, some visits may also have been recorded as Direct.

This means filtering GA4 for the AI Assistant channel alone won’t give you a complete historical view of traffic from AI platforms.

However, you can reconstruct some historical AI assistant traffic by looking for known AI platforms within your older session source data.

I’ve done this using Looker Studio and a calculated field that combines:

• traffic GA4 now classifies as AI Assistant
  
• historical sessions with identifiable AI assistant sources

It’s important to treat this as an approximation rather than a recreation of GA4’s own classification. It can only identify historical sessions where there is enough source information available to match the AI platform.

Here’s how I set it up.

Create A Calculated Field For AI Assistant Traffic In Looker Studio

In Looker Studio, create a new calculated field called:

AI Assistant Traffic

The aim is to identify sessions that either:

• GA4 already classifies within the AI Assistant channel, or
  
• have a session source matching one of the AI platforms you want to include historically

Use:

CASE

  WHEN REGEXP_MATCH(LOWER(Session default channel group), ".*ai assistant.*")

    OR REGEXP_MATCH(LOWER(Session source), ".*(chatgpt|openai|gemini|bard|perplexity|claude|anthropic|copilot|deepseek|grok).*")

  THEN "AI Assistant"

  ELSE "Other"

END

You can then use this calculated field as a filter within your Looker Studio report:

Set the filter to include:

AI Assistant Traffic = AI Assistant

This gives you a single reporting view combining newer sessions classified by GA4 as AI Assistant with older sessions where the source can be identified as one of the AI platforms in your regular expression.

The important distinction is that this doesn’t retrospectively change GA4’s historical channel classification. We’re using the historical source data that’s already available to create our own reporting layer in Looker Studio.

Group AI Assistant Sources Into Cleaner Names

Once you’ve identified the traffic, the next step is to make the report easier to read.

AI assistant traffic can appear under different source values. For example, ChatGPT traffic may contain different OpenAI-related sources, while Copilot traffic can appear under multiple Microsoft domains.

Rather than reporting each variation separately, I use another calculated field called:

AI Assistant Source

Use:

CASE
  WHEN REGEXP_MATCH(LOWER(Session source), ".*(chatgpt|chat\\.openai|openai).*")
    THEN "ChatGPT"

  WHEN REGEXP_MATCH(LOWER(Session source), ".*(gemini|bard).*")
    THEN "Gemini"

  WHEN REGEXP_MATCH(LOWER(Session source), ".*perplexity.*")
    THEN "Perplexity"

  WHEN REGEXP_MATCH(LOWER(Session source), ".*(claude|anthropic).*")
    THEN "Claude"

  WHEN REGEXP_MATCH(LOWER(Session source), ".*(copilot|copilot\\.microsoft|edgeservices).*")
    THEN "Copilot"

  WHEN REGEXP_MATCH(LOWER(Session source), ".*deepseek.*")
    THEN "DeepSeek"

  WHEN REGEXP_MATCH(LOWER(Session source), ".*grok.*")
    THEN "Grok"

  ELSE "Other"
END

You can then build your Looker Studio table using:

• Dimension: AI Assistant Source
  
• Metrics: Sessions, Users, Engagement rate, Key events and Revenue
  
• Filter: AI Assistant Traffic = AI Assistant

This gives you a much cleaner breakdown of which AI assistants are actually sending identifiable traffic, rather than having several source variations for the same platform.

One thing to bear in mind is that the matching rules aren’t permanent. AI platforms can change domains, referrer behaviour and how traffic is passed, so it’s worth periodically checking your raw Session source values and updating the expressions when necessary.

What This Report Can Show You

Once the report is set up, you can start looking beyond how much AI assistant traffic you’re receiving and understand what that traffic is actually doing on your website.

For example, you can use it to see:

• which AI assistants are sending identifiable traffic
  
• which landing pages receive that traffic
  
• whether AI assistant traffic is increasing over time
  
• how engagement compares with Organic Search and other channels
  
• which key events AI assistant visitors complete
  
• whether AI assistant traffic contributes to enquiries or revenue

This can be particularly useful when measuring your AI Optimisation and Visibility work.

AI assistant traffic may still be relatively small compared with established acquisition channels, so I wouldn’t judge its value purely on session volume. Looking at landing pages, engagement and conversions can give you a much better idea of whether that traffic is actually useful.

Here’s how I’ve brought those metrics together in my own report:

You can also view my full bespoke Looker Studio report.

Important Limitations

This method gives you a better view of AI assistant traffic, but it won’t identify every visit influenced by an AI platform.

The biggest limitation is referrer data. If someone clicks a link from an AI assistant and no usable referrer is passed, GA4 may record that session as Direct. There’s no reliable way to retrospectively identify those visits as AI assistant traffic.

There are a few other limitations worth keeping in mind:

• The historical report only identifies AI sources included in your matching rules.
  
• AI platforms can change domains or how they pass referral information.
  
• The calculated field is your own reporting classification, not GA4 retrospectively applying its AI Assistant channel.
  
• Visits from Google AI Overviews and AI Mode are different. Because these experiences are part of Google Search, that traffic is generally reported within Organic Search rather than the AI Assistant channel.

So I wouldn’t treat the resulting numbers as a complete measurement of every visit influenced by AI.

They’re better understood as identifiable AI assistant traffic based on the source and channel information available in GA4.

That distinction is important when reporting performance. If the report shows 100 AI Assistant sessions, for example, it doesn’t necessarily mean only 100 visits were influenced by AI. It means you’ve been able to identify 100 sessions using the available GA4 data and the matching rules you’ve created

How To Track AI Traffic in GA4 Going Forward

The new AI Assistant channel makes tracking identifiable AI traffic in GA4 much easier going forward.

For ongoing reporting, I’d use the native AI Assistant channel as the starting point. GA4 can then handle the classification when it recognises traffic from supported AI assistant sources.

For historical analysis, the Looker Studio method above lets you extend that view further back by identifying known AI platforms within your existing session source data.

Together, that gives you:

• GA4’s native AI Assistant classification for ongoing reporting
  
• a way to identify some historical AI referral traffic
  
• individual AI assistant reporting in Looker Studio
  
• landing page, engagement, key event and revenue analysis
  
• a reporting framework you can update as AI referral patterns change

It’s still important to remember that you’re measuring identifiable AI assistant traffic, not every visit that may have been influenced by AI.

But that’s a much more useful starting point than relying on the new AI Assistant channel alone, particularly if you want to understand how AI-referred traffic has changed over time.

For me, the real value isn’t simply knowing that AI assistants are sending traffic. It’s being able to see which platforms are sending it, where those visitors land, and whether they actually do anything useful once they arrive.

---