How to Track AI Search Referral Traffic in GA4

Traffic from AI-powered search tools is growing rapidly. ChatGPT, Perplexity, Google Gemini, Microsoft Copilot and other AI platforms now send measurable referral traffic to websites, and that traffic converts at rates that consistently beat traditional organic search. AI referrals to top websites hit 1.13 billion in June 2025, up 357% year-on-year (Similarweb). But GA4 doesn’t separate AI traffic by default. Without configuration, visits from ChatGPT and Perplexity get lumped into generic “Referral” traffic alongside everything else, making it invisible in standard reports.

At Gorilla Marketing, we configure analytics and tracking for clients as part of every SEO and AI optimisation engagement. AI traffic tracking is now standard in that setup. This guide covers how to identify, measure and report on AI search referral traffic in GA4, including the significant portion that current tools can’t track directly.

Why Track AI Search Traffic?

Three reasons this matters now, even though volume is still small for most sites.

It converts well. AI referral traffic consistently outperforms organic search. Seer Interactive found ChatGPT traffic converting at 15.9% compared to 1.76% for organic. Perplexity showed 10.5%, Claude 5%, Gemini 3%. A separate 13-month study from Search Engine Land found LLM traffic converting at approximately 18%, making it the highest-converting traffic source measured. The explanation: users arriving from AI tools have already had their basic questions answered and are clicking because they want depth, making them more qualified visitors.

It’s growing fast. AI platforms account for approximately 0.15% of global internet traffic, up 7x from 0.02% in 2024 (SE Ranking, study of 63,987 websites). UK AI traffic specifically tripled between January and April 2025. The volume is still roughly 25 times less than SEO or direct traffic, but the growth trajectory is steep. Setting up tracking now builds the baseline data you’ll need when this channel matters more.

It informs strategy. Knowing which pages receive AI traffic tells you which content is being cited by AI systems. That data directly feeds into your SEO and content strategy. Pages getting AI referrals are pages that AI systems trust and cite, and that’s valuable intelligence.

Which AI Platforms Send Referral Traffic?

ChatGPT dominates, accounting for around 84% of AI traffic in the UK specifically. But the long tail of smaller platforms is growing, and new tools launch regularly.

Important limitation: Google AI Overviews don’t send a separate referrer header. AI Overview clicks appear as standard Google organic traffic in GA4. There’s currently no reliable way to distinguish them within GA4 alone, though Google Search Console shows limited AI Overview impression data in the Search Appearance filter.

Setting Up a Custom Channel Group

The cleanest approach: create a custom channel group that automatically categorises AI traffic while keeping default groupings intact.

Step 1: Open channel groups.

In GA4, navigate to Admin > Data display > Channel groups. Select “Create new channel group.”

Step 2: Create the AI channel.

Name your group (e.g., “Custom with AI”). Add a new channel called “AI Search.”

Step 3: Set the matching condition.

Match session source using this regex:

Step 4: Set channel priority.

Move AI Search above the default Referral channel. GA4 processes rules sequentially, so the AI channel must be checked before traffic falls into the generic referral bucket.

Step 5: Save and apply.

The group applies from creation date forward. Custom channel groups don’t apply retroactively to historical data, so set this up now.

Viewing AI Traffic in Standard Reports

With the channel group configured, navigate to Reports > Acquisition > Traffic acquisition. Switch from “Default channel group” to your custom group. AI Search appears as its own row with sessions, engagement rate, conversions and other metrics.

Deeper Analysis with GA4 Explorations

For more granular analysis, GA4 Explorations let you segment AI traffic by landing page, device, geography and conversion events.

Create a new Exploration and add “Session source” as a dimension. Apply a filter using the same regex pattern. Add metrics: sessions, engaged sessions, key events, engagement rate, average engagement time.

Questions this answers:

Which pages receive the most AI referral traffic? These are the pages AI systems are citing.

Which AI platform sends the most engaged visitors? Platform quality varies significantly.

How does AI conversion rate compare to organic? Consistently higher, but the magnitude matters for budget decisions.

Are there geographic or device patterns? Mobile AI referrals behave differently from desktop.

For teams that want to go further, export GA4 data to BigQuery and use `REGEXP_CONTAINS` on the traffic source field to classify AI sessions. This enables unsampled historical analysis and custom attribution windows that GA4’s standard interface can’t support. It’s also the only way to do meaningful year-on-year comparison once you have enough data.

The Dark AI Traffic Problem

This is the biggest challenge with AI traffic measurement. The AI referrals you can see in GA4 are likely a fraction of actual AI-driven visits.

When someone copies a URL from a ChatGPT response and pastes it into their browser, GA4 records it as Direct. ChatGPT’s free tier rarely sends referrer headers. Mobile app referrals from virtually all AI platforms suppress referrer data entirely. Research from Loamly analysing 446,000 visits estimated that over 70% of AI-originated traffic gets misclassified, mostly appearing as Direct.

The referrer reliability varies significantly by platform:

ChatGPT began appending `utm_source=chatgpt.com` to citation links in June 2025, which improves tracking for those specific clicks. But this doesn’t apply to all link types, and sensitive topics (medical, legal, financial) have UTMs stripped.

Estimating Dark AI Traffic

No perfect solution exists, but several methods help estimate the hidden portion:

Behavioural filtering. In GA4, filter for sessions where the source is direct, the user is new, the landing page is a blog or informational page, and session duration exceeds three minutes. This segment likely contains hidden AI referrals, since these characteristics match AI visitor behaviour but not typical direct traffic.

Self-reported attribution. Add “How did you find us?” to sign-up or enquiry forms with “AI chatbot” as an option.

Server log analysis. Server logs sometimes reveal referrer strings from AI agents that client-side analytics miss.

Content-specific monitoring. Track Direct traffic to pages you know are frequently cited by AI tools. Unexplained spikes to those specific URLs often indicate unmeasured AI traffic.

What the Data Typically Shows

Across sites and industries, AI referral traffic shows consistent patterns:

Longer sessions. AI visitors average 9 minutes 19 seconds per session, 67.7% longer than organic search (SE Ranking). Claude referrals show the longest sessions at over 18 minutes. Perplexity and ChatGPT average around 9 minutes each.

Lower bounce rates. ChatGPT traffic bounces at approximately 35% compared to 48% for Google organic. Perplexity sits at 32%.

Higher conversion rates. Consistently above organic across every platform measured. The explanation is straightforward: AI visitors have already had their surface-level questions answered and are clicking through for depth, making them further along the decision journey.

Platform variation. ChatGPT and Perplexity dominate most referral mixes. Claude sends very little traffic but the visitors it does send are exceptionally engaged. The mix will shift as platforms evolve. A year ago, Perplexity barely registered. Now it accounts for 15% of AI referral traffic. Tracking by platform rather than as a single “AI” bucket captures these shifts.

How AI Consumption Differs from AI Referral

One important context for interpreting AI traffic data: AI systems consume far more content than they refer traffic back to. The crawl-to-refer ratio varies dramatically by platform. Claude crawls approximately 500,000 pages for every one referral it sends. ChatGPT’s ratio is roughly 3,700 to 1. Perplexity, which includes source links in every response, has the best ratio at approximately 700 to 1.

This means that even pages receiving zero AI referral traffic may be heavily used by AI systems as training or retrieval sources. AI traffic in GA4 represents the visible tip of a much larger iceberg of AI content consumption. Pages that appear frequently in AI responses (even without driving clicks) contribute to brand visibility and authority in ways that traditional traffic metrics don’t capture.

Connecting AI Traffic Data to Content Strategy

The practical value of AI traffic tracking goes beyond reporting. The data should feed directly into content decisions.

Identify citation-worthy content. Pages receiving AI referrals are pages AI systems trust enough to cite. Analyse what those pages have in common: clear definitional statements, original data, structured formatting, strong entity signals. These characteristics should inform how you structure new content.

Spot AI visibility gaps. If competitor pages are cited for queries your content should own, that’s a content gap. Monitor which competitor domains appear in AI responses for your target queries and compare against your own AI referral data.

Prioritise content updates. Pages receiving growing AI traffic are worth investing in. Update them with fresh data, improve their structure and ensure they maintain the qualities that earned the citations in the first place. Pages where AI traffic is declining may indicate that a competitor has published better content on the same topic.

Measure the ROI of AI optimisation. If you’re investing in content formats designed for AI citation, AI traffic data provides the clearest measure of whether that investment is working. Track AI referrals to pages created or updated specifically for AI visibility.

Reporting and Maintenance

Monthly review alongside standard channel reporting works for most sites. Key metrics to track:

Total AI sessions and month-on-month trend

AI as a percentage of total traffic (growth indicator)

Top landing pages from AI (which content is being cited)

Conversion rate by platform (traffic quality)

Engagement comparison (AI vs organic vs paid)

Review and update your regex patterns quarterly as new platforms emerge. ChatGPT, Perplexity and others change referrer behaviour without announcement. Adding a year suffix to your channel group name (e.g., “AI Traffic (2026)”) serves as a useful reminder to audit and update.

One practical note on GA4’s regex character limit: report-level filters cap regex at 250 characters. Use the full expanded pattern in custom channel groups (where there’s no limit), and a shorter version covering just the top three to four platforms for quick ad hoc report filters.

The Bigger Picture: AI Traffic in Context

AI referral traffic is still small relative to organic search. Google sends roughly 300 times more traffic than all AI platforms combined. The strategic importance isn’t in current volume but in trajectory and visitor quality.

A channel growing at 130% year-on-year in the UK, with conversion rates 4 to 10 times higher than organic, deserves measurement infrastructure even if it represents 0.1% of today’s traffic. Businesses that track this now will be the ones making data-informed decisions when it hits 5 to 10% of total traffic.

The zero-click search trend makes this doubly important. As more searches end without a click to any website, the clicks that do happen (including AI referrals) become more valuable per visit. AI traffic measurement isn’t a nice-to-have. It’s the foundation for understanding a channel that’s growing while others contract.

Gorilla Marketing includes AI traffic tracking in our analytics and tracking setup for all clients. If your analytics aren’t yet configured for AI referral measurement, set this up now while volumes are small so you have baseline data as the channel scales. Get in touch to discuss your analytics configuration.